{ "generated": "2026-10-09T04:33+08:00", "engines": [ { "engine": "llama.cpp", "repo": "ggml-org/llama.cpp", "role": "GGUF on anything: CUDA, ROCm, Vulkan, Metal, SYCL, CPU. Default for single-user boxes.", "version": "b11514", "date": "2026-10-08", "url": "https://github.com/ggml-org/llama.cpp/releases/tag/b11514" }, { "engine": "ik_llama.cpp", "repo": "ikawrakow/ik_llama.cpp", "role": "llama.cpp fork, faster CPU / hybrid MoE offload and IQ quants.", "version": "main@04b4ebc", "date": "2026-10-08", "url": "https://github.com/ikawrakow/ik_llama.cpp" }, { "engine": "vLLM", "repo": "vllm-project/vllm", "role": "Multi-user serving (PagedAttention, FP8 KV, MTP). NVIDIA, AMD Instinct, Intel XPU.", "version": "v0.31.0", "date": "2026-10-05", "url": "https://github.com/vllm-project/vllm/releases/tag/v0.31.0" }, { "engine": "SGLang", "repo": "sgl-project/sglang", "role": "Agentic / multi-turn serving (RadixAttention). NVIDIA, AMD Instinct.", "version": "v0.5.21", "date": "2026-10-02", "url": "https://github.com/sgl-project/sglang/releases/tag/v0.5.21" }, { "engine": "TensorRT-LLM", "repo": "NVIDIA/TensorRT-LLM", "role": "Max throughput on Hopper/Blackwell, native NVFP4.", "version": "v1.3.0rc29", "date": "2026-09-29", "url": "https://github.com/NVIDIA/TensorRT-LLM/releases/tag/v1.3.0rc29" }, { "engine": "Ollama", "repo": "ollama/ollama", "role": "Zero-config pull-and-run; easiest path for non-specialists.", "version": "v0.40.1", "date": "2026-10-07", "url": "https://github.com/ollama/ollama/releases/tag/v0.40.1" }, { "engine": "llama-swap", "repo": "mostlygeek/llama-swap", "role": "Hot-swap several models behind one OpenAI port.", "version": "v262", "date": "2026-10-03", "url": "https://github.com/mostlygeek/llama-swap/releases/tag/v262" }, { "engine": "MLX", "repo": "ml-explore/mlx", "role": "Apple Silicon array framework.", "version": "v0.32.3", "date": "2026-09-29", "url": "https://github.com/ml-explore/mlx/releases/tag/v0.32.3" }, { "engine": "mlx-lm", "repo": "ml-explore/mlx-lm", "role": "Apple Silicon LLM serving (mlx_lm.server is OpenAI-compatible).", "version": "v0.31.3", "date": "2026-04-22", "url": "https://github.com/ml-explore/mlx-lm/releases/tag/v0.31.3" }, { "engine": "mlx-vlm", "repo": "Blaizzy/mlx-vlm", "role": "Apple Silicon vision-language models.", "version": "v0.7.6", "date": "2026-10-05", "url": "https://github.com/Blaizzy/mlx-vlm/releases/tag/v0.7.6", "stale": true }, { "engine": "ExLlamaV3", "repo": "turboderp-org/exllamav3", "role": "Fastest single-stream on one NVIDIA GPU, EXL3 quants.", "version": "v1.6.0", "date": "2026-10-07", "url": "https://github.com/turboderp-org/exllamav3/releases/tag/v1.6.0" }, { "engine": "TabbyAPI", "repo": "theroyallab/tabbyAPI", "role": "OpenAI server for ExLlamaV3.", "version": "main@2fd6cc7", "date": "2026-10-06", "url": "https://github.com/theroyallab/tabbyAPI" }, { "engine": "KTransformers", "repo": "kvcache-ai/ktransformers", "role": "Huge MoE (DeepSeek-class) with CPU expert offload.", "version": "v0.7.1", "date": "2026-09-15", "url": "https://github.com/kvcache-ai/ktransformers/releases/tag/v0.7.1" }, { "engine": "Lemonade", "repo": "lemonade-sdk/lemonade", "role": "AMD Ryzen AI: NPU + iGPU, auto backend pick.", "version": "v2026.41.1", "date": "2026-10-07", "url": "https://github.com/lemonade-sdk/lemonade/releases/tag/v2026.41.1" }, { "engine": "OpenVINO", "repo": "openvinotoolkit/openvino", "role": "Intel CPU / Arc / NPU.", "version": "2026.4.1", "date": "2026-10-01", "url": "https://github.com/openvinotoolkit/openvino/releases/tag/2026.4.1" }, { "engine": "LocalAI", "repo": "mudler/LocalAI", "role": "All-in-one OpenAI-compatible server (LLM + image + audio).", "version": "v4.11.0", "date": "2026-10-02", "url": "https://github.com/mudler/LocalAI/releases/tag/v4.11.0" }, { "engine": "mistral.rs", "repo": "EricLBuehler/mistral.rs", "role": "Rust engine with in-situ quantization.", "version": "v0.9.4", "date": "2026-09-24", "url": "https://github.com/EricLBuehler/mistral.rs/releases/tag/v0.9.4" } ], "tokenmark": { "generated": "2026-10-07T21:14:44.466Z", "counts": { "configs": 258, "models": 118, "platforms": 4, "releases": 127 }, "platforms": [ "strix-halo", "dgx-spark", "mac-ultra", "mac-max" ], "configs": [ { "model": "Qwen3.6-35B-A3B-MTP", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "vulkan", "ctx": 262144, "mode": "mtp-3", "decode_tps": 66, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 41, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 39, "date": "" }, { "model": "Qwen3.8-27B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 131072, "mode": "mtp", "decode_tps": 24.6, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "UD-IQ2_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "vulkan", "ctx": 131072, "mode": null, "decode_tps": 12, "date": "" }, { "model": "Qwen3 0.6B", "quant": "Q8_0", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 266, "date": "" }, { "model": "Qwen3 0.6B", "quant": "Q8_0", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 208.7, "date": "" }, { "model": "LFM2.5 8B-A1B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 176.5, "date": "" }, { "model": "Qwen3-30B-A3B-Instruct-2507", "quant": "IQ4_XS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 103.2, "date": "" }, { "model": "Qwen3-Coder 30B-A3B", "quant": "Q4_K_S", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 98, "date": "" }, { "model": "Qwen3-Coder 30B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 97.1, "date": "" }, { "model": "Qwen3.6-35B-heretic", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 96, "date": "" }, { "model": "DeepSeek-V4-Flash 0731 1M [DUAL 2xSpark]", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 95.5, "date": "" }, { "model": "Qwen3.6-35B-A3B-NVFP4-Fast (Unsloth)", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 92.6, "date": "" }, { "model": "GPT-OSS-20B", "quant": "Q4_K_XL", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 90.5, "date": "" }, { "model": "Qwen3-Coder 30B-A3B", "quant": "IQ4_XS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 90.4, "date": "" }, { "model": "Qwen3 30B-A3B NEO-MAX", "quant": "IQ4_XS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 87.4, "date": "" }, { "model": "Ornith-1.0-35B-AEON", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 84.8, "date": "" }, { "model": "Qwen3.6 35B-A3B", "quant": "Q4_0", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 81.3, "date": "" }, { "model": "Qwen3-30B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 79.1, "date": "" }, { "model": "Nemotron Cascade 2 30B-A3B", "quant": "IQ4_XS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 79, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 77.4, "date": "" }, { "model": "Nemotron 3 Nano 30B-A3B", "quant": "IQ4_XS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 76, "date": "" }, { "model": "Qwen3.5 35B-A3B", "quant": "IQ4_XS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 75.2, "date": "" }, { "model": "Gemma 4 26B-A4B IT QAT", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 74.8, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-3", "decode_tps": 74.6, "date": "" }, { "model": "Qwen3-Coder 30B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 73.7, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-2", "decode_tps": 72.8, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-3", "decode_tps": 68.3, "date": "" }, { "model": "Qwen AgentWorld 35B-A3B", "quant": "IQ4_XS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 65.7, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-3", "decode_tps": 65.6, "date": "" }, { "model": "Qwen3.5 35B-A3B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 64.9, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-2", "decode_tps": 64.5, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-2", "decode_tps": 64.4, "date": "" }, { "model": "Nemotron 3 Nano Omni 30B-A3B Reasoning", "quant": "MXFP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 64.3, "date": "" }, { "model": "Qwen3.6 35B-A3B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 63.8, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 62.8, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "Q4_K_XL", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 62.7, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 62.2, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-3", "decode_tps": 62.1, "date": "" }, { "model": "Qwen3-Coder-Next 80B-A3B", "quant": "IQ4_XS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 61.9, "date": "" }, { "model": "Nemotron Labs Audex 30B-A3B text-only", "quant": "MXFP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 60.7, "date": "" }, { "model": "Qwen3.6 35B-A3B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "ollama", "ctx": 2048, "mode": null, "decode_tps": 60.6, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-2", "decode_tps": 58.9, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "baseline", "decode_tps": 58.7, "date": "" }, { "model": "NVIDIA Nemotron 3 Nano Omni 30B-A3B Reasoning", "quant": "MXFP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 56.6, "date": "" }, { "model": "Ornith-1.0-35B", "quant": "Q8_0", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 56.1, "date": "" }, { "model": "Qwen3-Next 80B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 54.9, "date": "" }, { "model": "Qwen3.5 35B-A3B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 54.7, "date": "" }, { "model": "Gemma 4 26B-A4B IT", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 54.2, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "baseline", "decode_tps": 54.1, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 53.5, "date": "" }, { "model": "Nemotron 3 Nano Omni 30B-A3B", "quant": "NVFP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 53.2, "date": "" }, { "model": "llama-2-7b", "quant": "Q4_0", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 53, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 53, "date": "" }, { "model": "Qwen3.6 35B-A3B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 52.7, "date": "" }, { "model": "DeepSeek-V4-Flash DSpark 1M [DUAL 2xSpark]", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 52.6, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 51.7, "date": "" }, { "model": "gpt-oss-120b", "quant": "MXFP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 50.6, "date": "" }, { "model": "llama-2-7b", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 50.4, "date": "" }, { "model": "Qwen3-Coder-Next-80B-A3B", "quant": "Q4_K_XL", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 50.1, "date": "" }, { "model": "Qwen3-Next 80B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 49.6, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 49.2, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 48.9, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "baseline", "decode_tps": 48.7, "date": "" }, { "model": "Gemma 4 26B-A4B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 48.5, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": null, "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 48.2, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 48.1, "date": "" }, { "model": "Nemotron-3-Nano-30B-A3B", "quant": null, "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 47.9, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 47.9, "date": "" }, { "model": "gpt-oss-20b-F16", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 47.8, "date": "" }, { "model": "Kimi-Linear-48B-A3B", "quant": "Q8_0", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 47.2, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 46.5, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "baseline", "decode_tps": 46, "date": "" }, { "model": "shisa-v2-llama3.1-8b.i1", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 43.5, "date": "" }, { "model": "gemma-4-26B-A4B-it", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 41.8, "date": "" }, { "model": "Nemotron-3-Nano-Omni-30B-A3B", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 41.7, "date": "" }, { "model": "Nemotron-3-Nano-30B-A3B", "quant": "FP8", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 40.7, "date": "" }, { "model": "Qwen3.6-27B", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": "mtp-3", "decode_tps": 37.3, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "UD-IQ3_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 524288, "mode": "dspark", "decode_tps": 35.7, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 35.1, "date": "" }, { "model": "gpt-oss-120b-F16", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 33.9, "date": "" }, { "model": "Nemotron-Labs-3-Puzzle-75B-A9B", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 32.8, "date": "" }, { "model": "DeepSeek-V4-Flash (0731)", "quant": null, "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 32.8, "date": "" }, { "model": "Kolibri-1", "quant": "FP8", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 32, "date": "" }, { "model": "MiniMax-M3 428B [DUAL 2xSpark]", "quant": "GPTQ", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 30.2, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "rocm", "ctx": 131072, "mode": "dspark", "decode_tps": 29.6, "date": "" }, { "model": "Gemma 4 12B IT QAT", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 29.3, "date": "" }, { "model": "Gemma-4-12B-IT", "quant": "Q4_K_M", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 28.4, "date": "" }, { "model": "Qwen3.6-27B-AEON", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": "dflash-10", "decode_tps": 27.9, "date": "" }, { "model": "Qwen3.8-27B", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 65536, "mode": "main-low", "decode_tps": 27.8, "date": "" }, { "model": "Laguna-S-2.1", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 27.3, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "IQ4_XS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 27.2, "date": "" }, { "model": "Qwen-AgentWorld-35B-A3B", "quant": "BF16", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 26.9, "date": "" }, { "model": "Laguna-S-2.1 [nspec=15]", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 25.9, "date": "" }, { "model": "GLM-5.3-Flash", "quant": null, "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 25.1, "date": "" }, { "model": "Nemotron-3-Super-120B-A12B", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 24.5, "date": "" }, { "model": "Qwen3.8-27B", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 65536, "mode": "main-off", "decode_tps": 23.9, "date": "" }, { "model": "GLM-4.5-Air", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 23.4, "date": "" }, { "model": "Qwen3.6-27B-NVFP4 (Unsloth)", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": "mtp-2", "decode_tps": 23.3, "date": "" }, { "model": "Qwen3.8-27B", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "sglang", "ctx": null, "mode": "dspark-8", "decode_tps": 22.8, "date": "" }, { "model": "dots.llm1.inst", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 22.7, "date": "" }, { "model": "Qwen3.8-27B", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 65536, "mode": "main-medium", "decode_tps": 22.6, "date": "" }, { "model": "Gemma-4-26B-A4B-IT", "quant": "BF16", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 22.3, "date": "" }, { "model": "Qwen3.5-122B-A10B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 21.6, "date": "" }, { "model": "MiMo-V2.5 Omni [DUAL 2xSpark]", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 21.5, "date": "" }, { "model": "Qwen3.8 27B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "ollama", "ctx": 4096, "mode": null, "decode_tps": 20.4, "date": "" }, { "model": "Llama-4-Scout-17B-16E-Instruct", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 20.2, "date": "" }, { "model": "Step-3.7-Flash", "quant": "IQ4_XS", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 19.9, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "UD-IQ3_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 19.8, "date": "" }, { "model": "Nex-N2-Pro-397B-A17B", "quant": null, "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 18.9, "date": "" }, { "model": "Nemotron 3 Super 120B-A12B", "quant": "IQ4_XS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 18.9, "date": "" }, { "model": "Hunyuan-A13B-Instruct", "quant": "UD-Q6_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 18.4, "date": "" }, { "model": "Llama 4 Scout 109B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 18.3, "date": "" }, { "model": "SuperQwen3.8-27B-abliterated", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 17.3, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "UD-IQ2_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 16.2, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "UD-IQ2_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 16.1, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "UD-IQ3_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 16, "date": "" }, { "model": "Qwen3.8-27B", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 131072, "mode": "main-low", "decode_tps": 16, "date": "" }, { "model": "Qwen3-235B-A22B-Instruct-2507", "quant": "UD-Q3_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 15.9, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "UD-IQ3_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 15.8, "date": "" }, { "model": "Qwen3.8-27B", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 131072, "mode": "main-medium", "decode_tps": 15.6, "date": "" }, { "model": "Qwen3.8-27B", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 131072, "mode": "main-xhigh", "decode_tps": 15, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "UD-IQ2_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 14.8, "date": "" }, { "model": "Mistral-Small-3.1-24B-Instruct-2503", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 14.7, "date": "" }, { "model": "Qwen3.8-27B", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 131072, "mode": "main-off", "decode_tps": 14.7, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "UD-IQ3_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 14.5, "date": "" }, { "model": "Qwen3.8-27B", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 65536, "mode": "variance-low", "decode_tps": 14, "date": "" }, { "model": "Qwen3.5-397B-A17B [DUAL 2xSpark]", "quant": "IQ4_NL", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 13.6, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-3", "decode_tps": 13.5, "date": "" }, { "model": "Qwen3.8-27B", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 65536, "mode": "variance-medium", "decode_tps": 13.5, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-3", "decode_tps": 13.3, "date": "" }, { "model": "DeepSeek V4 Flash 284B", "quant": "UD-IQ2_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 13.3, "date": "" }, { "model": "Qwen3.8-27B", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 65536, "mode": "main-xhigh", "decode_tps": 13.3, "date": "" }, { "model": "Qwen3.6 27B MTP NVFP4 v3", "quant": "NVFP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 13.2, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "UD-IQ2_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 13.1, "date": "" }, { "model": "Qwen3.8-27B", "quant": null, "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 65536, "mode": "budget-medium", "decode_tps": 12.9, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-3", "decode_tps": 12.8, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-2", "decode_tps": 12.4, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-3", "decode_tps": 12.2, "date": "" }, { "model": "gemma-3-27b-it", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 12.1, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-2", "decode_tps": 11.7, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-2", "decode_tps": 11.7, "date": "" }, { "model": "Gemma 4 31B IT QAT", "quant": "Q4_0", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 11.4, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp-2", "decode_tps": 10.8, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "UD-IQ2_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 9.1, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "UD-IQ3_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 9, "date": "" }, { "model": "Qwen3.6-27B", "quant": "Q8_0", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 7.8, "date": "" }, { "model": "Qwen3.6 27B MTP", "quant": "Q8_0", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 7.7, "date": "" }, { "model": "Qwopus3.6-27B-Coder", "quant": "Q8_0", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 7.6, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 6.6, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 6.6, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 6.6, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 6.6, "date": "" }, { "model": "Gemma-4-31B-IT", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 6.6, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "baseline", "decode_tps": 6.5, "date": "" }, { "model": "Qwen3-32B", "quant": "Q8_0", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 6.4, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "baseline", "decode_tps": 6.3, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "baseline", "decode_tps": 6.3, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q8_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "baseline", "decode_tps": 6.1, "date": "" }, { "model": "shisa-v2-llama3.3-70b.i1", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 5.1, "date": "" }, { "model": "Llama 3.1 70B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "ollama", "ctx": null, "mode": null, "decode_tps": 4.7, "date": "" }, { "model": "Ornith-1.5-35B-A3B", "quant": "FP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": null, "ctx": null, "mode": "mtp", "decode_tps": 105.6, "date": "" }, { "model": "NVIDIA Nemotron 3.5 Lightning 30B", "quant": "FP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": null, "ctx": null, "mode": "mtp", "decode_tps": 95.2, "date": "" }, { "model": "Gemma-4-26B-A4B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 78, "date": "" }, { "model": "Nex N2.5 Mini 35B-A3B", "quant": "FP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": null, "ctx": null, "mode": null, "decode_tps": 76.9, "date": "" }, { "model": "Ornith 1.5 35B-A3B", "quant": "FP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": null, "ctx": null, "mode": null, "decode_tps": 76.9, "date": "" }, { "model": "Nex N2.5 Mini 35B-A3B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": null, "ctx": null, "mode": null, "decode_tps": 72.8, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 262144, "mode": null, "decode_tps": 72, "date": "" }, { "model": "Ornith 1.5 35B-A3B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": null, "ctx": null, "mode": null, "decode_tps": 71.6, "date": "" }, { "model": "GLM-5.3-Flash", "quant": null, "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": 262144, "mode": null, "decode_tps": 38, "date": "" }, { "model": "Qwen 3.8 27B", "quant": "FP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": null, "ctx": null, "mode": "mtp-4", "decode_tps": 33.8, "date": "" }, { "model": "DeepSeek V4 Flash 284B", "quant": "IQ2_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": null, "ctx": null, "mode": "mtp", "decode_tps": 32, "date": "" }, { "model": "Qwen 3.8 27B", "quant": "FP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": null, "ctx": 32768, "mode": "mtp", "decode_tps": 26.9, "date": "" }, { "model": "Muse-Glimmer-30B", "quant": "FP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "rocm", "ctx": null, "mode": "dflash", "decode_tps": 26.5, "date": "" }, { "model": "Qwen3.8-27B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "vulkan", "ctx": 262144, "mode": "mtp", "decode_tps": 23.5, "date": "" }, { "model": "Qwen3.8-27B VL", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp", "decode_tps": 22, "date": "" }, { "model": "Qwen3.8-27B-DFlash2", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "vulkan", "ctx": 32768, "mode": "dflash2", "decode_tps": 21.2, "date": "" }, { "model": "Qwen3.6-27B", "quant": "UD-Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 262144, "mode": "mtp-5", "decode_tps": 21, "date": "" }, { "model": "Qwen 3.8 27B", "quant": "FP8", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": "mtp", "decode_tps": 19, "date": "" }, { "model": "Qwen3.8-27B", "quant": "Q8_0", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "vulkan", "ctx": 131072, "mode": "mtp", "decode_tps": 17, "date": "" }, { "model": "Qwen 3.8 27B", "quant": "FP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "vulkan", "ctx": 16384, "mode": null, "decode_tps": 13.9, "date": "" }, { "model": "Qwen 3.8 27B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": null, "ctx": null, "mode": null, "decode_tps": 12.4, "date": "" }, { "model": "Qwen 3.8 27B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 32768, "mode": null, "decode_tps": 12.3, "date": "" }, { "model": "Qwen 3.8 27B", "quant": "FP16", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 5, "date": "" }, { "model": "llama-3.2-1b", "quant": "MLX-4BIT", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": 512, "mode": null, "decode_tps": 527.7, "date": "" }, { "model": "llama-3.2-1b", "quant": "Q4_K_M", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 439.6, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "sglang", "ctx": 262144, "mode": "c6", "decode_tps": 306, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "sglang", "ctx": 262144, "mode": null, "decode_tps": 275.4, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "FP8", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": null, "ctx": null, "mode": null, "decode_tps": 266.8, "date": "" }, { "model": "qwen3-4b", "quant": "MLX-4BIT", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": 512, "mode": null, "decode_tps": 186, "date": "" }, { "model": "qwen3-4b", "quant": "Q4_K_M", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 159.7, "date": "" }, { "model": "qwen3.6-35b-a3b", "quant": "MLX-4BIT", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": 512, "mode": null, "decode_tps": 151.1, "date": "" }, { "model": "qwen3-30b-a3b", "quant": "Q4_K_M", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 148.5, "date": "" }, { "model": "qwen3-30b-a3b", "quant": "MLX-4BIT", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": 512, "mode": null, "decode_tps": 147.2, "date": "" }, { "model": "Qwen3 1.7B", "quant": "Q4_K_M", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": 2048, "mode": null, "decode_tps": 146.1, "date": "" }, { "model": "qwen3.6-35b-a3b", "quant": "Q4_K_M", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 124.3, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "4-bit", "hardware": "Apple Mac Studio (M-series Ultra)", "hw": "mac-ultra", "backend": "mlx", "ctx": 13900, "mode": "batch8", "decode_tps": 112, "date": "" }, { "model": "qwen3-8b", "quant": "MLX-4BIT", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": 512, "mode": null, "decode_tps": 110.5, "date": "" }, { "model": "Qwen3.6 35B-A3B", "quant": "MLX-4BIT", "hardware": "Apple Mac Studio (M-series Ultra)", "hw": "mac-ultra", "backend": "mlx", "ctx": 32000, "mode": null, "decode_tps": 109.1, "date": "" }, { "model": "qwen3-8b", "quant": "Q4_K_M", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 97.7, "date": "" }, { "model": "Ministral 3B", "quant": "Q4_K_M", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": 2048, "mode": null, "decode_tps": 86.6, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "FP8", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "metrale", "ctx": 2048, "mode": "mtp-1", "decode_tps": 85, "date": "" }, { "model": "DeepSeek-V4-Flash-DSpark", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": 1000000, "mode": "peak", "decode_tps": 84.3, "date": "" }, { "model": "Qwen3 30B (MoE)", "quant": "Q4_K_M", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": 2048, "mode": null, "decode_tps": 83.8, "date": "" }, { "model": "GLM-5.3-Flash", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "sglang", "ctx": 131072, "mode": "dflash2-c12", "decode_tps": 83.2, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "MLX-4BIT", "hardware": "Apple Mac Studio (M-series Ultra)", "hw": "mac-ultra", "backend": "mlx", "ctx": null, "mode": "mtp", "decode_tps": 76.4, "date": "" }, { "model": "gemma-3-12b", "quant": "MLX-4BIT", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": 512, "mode": null, "decode_tps": 66.9, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "sglang", "ctx": 262144, "mode": "c1", "decode_tps": 63, "date": "" }, { "model": "qwen3-14b", "quant": "MLX-4BIT", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": 512, "mode": null, "decode_tps": 62.5, "date": "" }, { "model": "gemma-3-12b", "quant": "Q4_K_M", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 61.5, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": null, "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "cuda", "ctx": null, "mode": null, "decode_tps": 59, "date": "" }, { "model": "qwen3-14b", "quant": "Q4_K_M", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 57.4, "date": "" }, { "model": "Qwen3 14B", "quant": "MLX-4BIT", "hardware": "Apple Mac Studio (M-series Ultra)", "hw": "mac-ultra", "backend": "mlx", "ctx": 32000, "mode": null, "decode_tps": 55.4, "date": "" }, { "model": "DeepSeek-V4-Flash-0731-JA-REAP-K216", "quant": "EXL3-3BPW", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": null, "ctx": 256000, "mode": null, "decode_tps": 55, "date": "" }, { "model": "gpt-oss-120b", "quant": "MXFP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 128, "mode": null, "decode_tps": 52.9, "date": "" }, { "model": "gpt-oss-120b", "quant": "MXFP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 128, "mode": null, "decode_tps": 50.2, "date": "" }, { "model": "Qwen3-Coder-Next", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": null, "ctx": 64000, "mode": null, "decode_tps": 50, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "W4B", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "halogen", "ctx": 32768, "mode": "mtp", "decode_tps": 46, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "4-bit", "hardware": "Apple Mac Studio (M-series Ultra)", "hw": "mac-ultra", "backend": "mlx", "ctx": 13900, "mode": "single", "decode_tps": 45, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": 262144, "mode": "mtp-2", "decode_tps": 44.2, "date": "" }, { "model": "Qwen3.8 27B", "quant": "MLX-4BIT", "hardware": "Apple Mac Studio (M-series Ultra)", "hw": "mac-ultra", "backend": "mlx", "ctx": 32000, "mode": null, "decode_tps": 42.6, "date": "" }, { "model": "Qwen3 8B", "quant": "Q4_K_M", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": 2048, "mode": null, "decode_tps": 42, "date": "" }, { "model": "mistral-small-3.1-24b", "quant": "MLX-4BIT", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": 512, "mode": null, "decode_tps": 41.1, "date": "" }, { "model": "Ministral 8B", "quant": "Q4_K_M", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": 2048, "mode": null, "decode_tps": 40.1, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "W4B", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "halogen", "ctx": 1500, "mode": "baseline", "decode_tps": 37.6, "date": "" }, { "model": "mistral-small-3.1-24b", "quant": "Q4_K_M", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 37.2, "date": "" }, { "model": "Qwen3.8-Flash-Next-REAP-288", "quant": "MLX-4BIT", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": null, "mode": null, "decode_tps": 37, "date": "" }, { "model": "GLM-5.3-Flash", "quant": "FP8", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": null, "decode_tps": 36, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "IQ3_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 156000, "mode": "mtp", "decode_tps": 35.7, "date": "" }, { "model": "GLM-5.3-Flash", "quant": "MLX-4BIT", "hardware": "Apple Mac Studio (M-series Ultra)", "hw": "mac-ultra", "backend": "mlx", "ctx": null, "mode": null, "decode_tps": 34.2, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "2.58bpw-mix", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": null, "ctx": null, "mode": "dspark", "decode_tps": 34.2, "date": "" }, { "model": "DeepSeek-V4-Flash-0731", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": 1000000, "mode": "prose", "decode_tps": 33.2, "date": "" }, { "model": "gemma-3-27b", "quant": "MLX-4BIT", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": 512, "mode": null, "decode_tps": 31.8, "date": "" }, { "model": "Qwen3.6-35B-A3B", "quant": "BF16", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": null, "mode": null, "decode_tps": 30.8, "date": "" }, { "model": "gemma-3-27b", "quant": "Q4_K_M", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 29.9, "date": "" }, { "model": "DeepSeek-V4-Flash-0731-MXFP4-MLX-Abliterated", "quant": "MXFP4", "hardware": "Apple Mac Studio (M-series Ultra)", "hw": "mac-ultra", "backend": "mlx", "ctx": 1048576, "mode": null, "decode_tps": 29.6, "date": "" }, { "model": "GLM-5.3-Flash", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "sglang", "ctx": 131072, "mode": "dflash2-c1", "decode_tps": 28.6, "date": "" }, { "model": "qwen3-32b", "quant": "MLX-4BIT", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": 512, "mode": null, "decode_tps": 28.4, "date": "" }, { "model": "Qwen3.8-Flash-Next-REAP-288", "quant": "MLX-4BIT", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "mlx", "ctx": null, "mode": null, "decode_tps": 28, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": null, "mode": "mtp-1", "decode_tps": 27.6, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": 262144, "mode": "baseline", "decode_tps": 27.3, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": 1000000, "mode": null, "decode_tps": 26.7, "date": "" }, { "model": "qwen3-32b", "quant": "Q4_K_M", "hardware": "Apple M-series Max (MacBook Pro / Mac Studio Max)", "hw": "mac-max", "backend": "llama.cpp", "ctx": 512, "mode": null, "decode_tps": 26.4, "date": "" }, { "model": "SuperQwen3.8-27b-abliterated", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "vllm", "ctx": 262043, "mode": "spec-k5", "decode_tps": 25.8, "date": "" }, { "model": "Ministral 14B", "quant": "Q4_K_M", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": 2048, "mode": null, "decode_tps": 25.7, "date": "" }, { "model": "Qwen3.8-27B", "quant": "NVFP4", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "metrale", "ctx": 2048, "mode": "mtp", "decode_tps": 25.1, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "Q4_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 24, "date": "" }, { "model": "GLM-5.3-Flash Abliterated", "quant": "MLX-4BIT", "hardware": "Apple Mac Studio (M-series Ultra)", "hw": "mac-ultra", "backend": "mlx", "ctx": 16384, "mode": "mtp", "decode_tps": 24, "date": "" }, { "model": "Qwen3.8-Flash-Next", "quant": "IQ3_XXS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 156000, "mode": "baseline", "decode_tps": 23.5, "date": "" }, { "model": "Qwen3.8-27B", "quant": "UD-Q6_K_XL", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 131072, "mode": "dflash2", "decode_tps": 17, "date": "" }, { "model": "Motif-3 315B", "quant": "UD-IQ2_XXS", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": 32768, "mode": null, "decode_tps": 16.5, "date": "" }, { "model": "SuperQwen3.8-Flash-Next-abliterated", "quant": "FP8", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "sglang", "ctx": 8192, "mode": null, "decode_tps": 16.2, "date": "" }, { "model": "Qwen3.8-27B", "quant": "IQ4_XS", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": 65536, "mode": null, "decode_tps": 14.1, "date": "" }, { "model": "Qwen 3.8 27B", "quant": "FP4", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 14, "date": "" }, { "model": "Qwen 3.8 27B", "quant": "Q4_K_M", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 12.3, "date": "" }, { "model": "Qwen3 32B (Dense)", "quant": "Q4_K_M", "hardware": "NVIDIA DGX Spark (GB10)", "hw": "dgx-spark", "backend": "llama.cpp", "ctx": 2048, "mode": null, "decode_tps": 10.5, "date": "" }, { "model": "Qwen 3.8 27B", "quant": "FP8", "hardware": "AMD Strix Halo (Ryzen AI Max+ 395)", "hw": "strix-halo", "backend": "llama.cpp", "ctx": null, "mode": null, "decode_tps": 7.7, "date": "" } ] }, "hf_trending": [ { "id": "ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF", "format": "GGUF", "created": "2026-09-07", "downloads": 3405442, "url": "https://huggingface.co/ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF" }, { "id": "unsloth/embeddinggemma-2-GGUF", "format": "GGUF", "created": "2026-10-06", "downloads": 29692, "url": "https://huggingface.co/unsloth/embeddinggemma-2-GGUF" }, { "id": "ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF", "format": "GGUF", "created": "2026-08-28", "downloads": 1517150, "url": "https://huggingface.co/ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF" }, { "id": "unsloth/Qwen3.8-27B-GGUF", "format": "GGUF", "created": "2026-08-13", "downloads": 6518857, "url": "https://huggingface.co/unsloth/Qwen3.8-27B-GGUF" }, { "id": "ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-Coder-GGUF", "format": "GGUF", "created": "2026-09-26", "downloads": 599869, "url": "https://huggingface.co/ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-Coder-GGUF" }, { "id": "nvidia/Nemotron-3-Diarization", "format": "GGUF", "created": "2026-09-01", "downloads": 66040, "url": "https://huggingface.co/nvidia/Nemotron-3-Diarization" }, { "id": "unsloth/Qwen3.8-Flash-Next-GGUF", "format": "GGUF", "created": "2026-08-26", "downloads": 1315187, "url": "https://huggingface.co/unsloth/Qwen3.8-Flash-Next-GGUF" }, { "id": "unsloth/Qwen-Image-2.1-GGUF", "format": "GGUF", "created": "2026-09-21", "downloads": 805462, "url": "https://huggingface.co/unsloth/Qwen-Image-2.1-GGUF" }, { "id": "ggml-org/embeddinggemma-2-GGUF", "format": "GGUF", "created": "2026-10-06", "downloads": 9518, "url": "https://huggingface.co/ggml-org/embeddinggemma-2-GGUF" }, { "id": "LiquidAI/d1-3B-GGUF", "format": "GGUF", "created": "2026-10-06", "downloads": 2353, "url": "https://huggingface.co/LiquidAI/d1-3B-GGUF" }, { "id": "ggml-org/Laya-GGUF", "format": "GGUF", "created": "2026-10-01", "downloads": 10112, "url": "https://huggingface.co/ggml-org/Laya-GGUF" }, { "id": "nvidia/nemotron-3.5-asr-streaming-0.6b", "format": "GGUF", "created": "2026-05-15", "downloads": 1289845, "url": "https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b" }, { "id": "LiquidAI/d1-omni-600M-GGUF", "format": "GGUF", "created": "2026-10-06", "downloads": 826, "url": "https://huggingface.co/LiquidAI/d1-omni-600M-GGUF" }, { "id": "bartowski/MiMo-V2.6-Distill-Qwen-9B-GGUF", "format": "GGUF", "created": "2026-09-21", "downloads": 496776, "url": "https://huggingface.co/bartowski/MiMo-V2.6-Distill-Qwen-9B-GGUF" }, { "id": "ggml-org/Clef-Flash-GGUF", "format": "GGUF", "created": "2026-10-02", "downloads": 26306, "url": "https://huggingface.co/ggml-org/Clef-Flash-GGUF" }, { "id": "unsloth/gemma-4-26B-A4B-it-GGUF", "format": "GGUF", "created": "2026-04-01", "downloads": 565548, "url": "https://huggingface.co/unsloth/gemma-4-26B-A4B-it-GGUF" }, { "id": "mlx-community/clef-flash-4bit", "format": "MLX", "created": "2026-10-01", "downloads": 3804, "url": "https://huggingface.co/mlx-community/clef-flash-4bit" }, { "id": "mlx-community/Qwen3.8-27B-4bit", "format": "MLX", "created": "2026-08-14", "downloads": 429036, "url": "https://huggingface.co/mlx-community/Qwen3.8-27B-4bit" }, { "id": "mlx-community/clef-4bit", "format": "MLX", "created": "2026-10-01", "downloads": 1355, "url": "https://huggingface.co/mlx-community/clef-4bit" }, { "id": "mlx-community/clef-8bit", "format": "MLX", "created": "2026-10-02", "downloads": 1085, "url": "https://huggingface.co/mlx-community/clef-8bit" }, { "id": "mlx-community/embeddinggemma-2-bf16", "format": "MLX", "created": "2026-10-06", "downloads": 1001, "url": "https://huggingface.co/mlx-community/embeddinggemma-2-bf16" }, { "id": "mlx-community/Llama-3-Groq-8B-Tool-Use-4bit", "format": "MLX", "created": "2024-07-17", "downloads": 603, "url": "https://huggingface.co/mlx-community/Llama-3-Groq-8B-Tool-Use-4bit" }, { "id": "mlx-community/Qwen3.5-9B-MLX-4bit", "format": "MLX", "created": "2026-03-02", "downloads": 31373, "url": "https://huggingface.co/mlx-community/Qwen3.5-9B-MLX-4bit" }, { "id": "mlx-community/Qwen3.8-27B-8bit", "format": "MLX", "created": "2026-08-14", "downloads": 78188, "url": "https://huggingface.co/mlx-community/Qwen3.8-27B-8bit" } ], "models": [ { "name": "Qwen3.8-27B", "params_b": 27.3, "architecture": "dense", "active_params_b": null, "capabilities": [ "coding", "vision", "chat", "general" ], "gguf_repo": "unsloth/Qwen3.8-27B-GGUF", "mmproj": true, "quants": [ { "name": "UD-Q2_K_XL", "size_mb": 9829, "bpw": 2.88 }, { "name": "UD-Q3_K_XL", "size_mb": 13146, "bpw": 3.85 }, { "name": "UD-IQ4_XS", "size_mb": 14253, "bpw": 4.17 }, { "name": "UD-Q4_K_XL", "size_mb": 17559, "bpw": 5.14 }, { "name": "UD-Q5_K_XL", "size_mb": 20877, "bpw": 6.11 }, { "name": "UD-Q6_K_XL", "size_mb": 25299, "bpw": 7.41 }, { "name": "Q8_0", "size_mb": 29047, "bpw": 8.51 }, { "name": "UD-Q8_K_XL", "size_mb": 31458, "bpw": 9.21 } ], "notes": "Dense 27B with native vision (needs the mmproj file). Strong all-rounder that fits one 24-32 GB GPU at 4-bit.", "source": "https://huggingface.co/Qwen/Qwen3.8-27B", "fetched": "2026-10-09T04:33+08:00" }, { "name": "Qwen3.8-Flash-Next", "params_b": 176.9, "architecture": "moe", "active_params_b": 6, "capabilities": [ "coding", "vision", "chat", "general" ], "gguf_repo": "unsloth/Qwen3.8-Flash-Next-GGUF", "mmproj": true, "quants": [ { "name": "UD-Q2_K_XL", "size_mb": 78869, "bpw": 3.57 }, { "name": "UD-Q3_K_XL", "size_mb": 89986, "bpw": 4.07 }, { "name": "UD-IQ4_XS", "size_mb": 93683, "bpw": 4.24 }, { "name": "UD-Q4_K_XL", "size_mb": 111335, "bpw": 5.03 }, { "name": "UD-Q5_K_XL", "size_mb": 158286, "bpw": 7.16 }, { "name": "UD-Q6_K_XL", "size_mb": 169165, "bpw": 7.65 }, { "name": "Q8_0", "size_mb": 188225, "bpw": 8.51 } ], "notes": "MoE, ~6B active of 125B plus 51B n-gram embedding. Native vision (mmproj). MTP draft files in the repo's MTP/ folder speed up decode.", "source": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next", "fetched": "2026-10-09T04:33+08:00" }, { "name": "GLM-5.3-Flash", "params_b": 320.8, "architecture": "moe", "active_params_b": 18, "capabilities": [ "coding", "chat", "general", "vision" ], "gguf_repo": "unsloth/GLM-5.3-Flash-GGUF", "mmproj": true, "quants": [ { "name": "UD-Q2_K_XL", "size_mb": 108720, "bpw": 2.71 }, { "name": "UD-Q3_K_XL", "size_mb": 147536, "bpw": 3.68 }, { "name": "UD-IQ4_XS", "size_mb": 156822, "bpw": 3.91 }, { "name": "UD-Q4_K_XL", "size_mb": 199707, "bpw": 4.98 }, { "name": "UD-Q5_K_XL", "size_mb": 240306, "bpw": 5.99 }, { "name": "UD-Q6_K_XL", "size_mb": 291833, "bpw": 7.28 }, { "name": "Q8_0", "size_mb": 340982, "bpw": 8.5 } ], "notes": "MoE, 320B total, 18B active. Needs 128 GB+ even at 2-bit.", "source": "https://huggingface.co/zai-org/GLM-5.3-Flash", "fetched": "2026-10-09T04:33+08:00" }, { "name": "gpt-oss-120b", "params_b": 116.8, "architecture": "moe", "active_params_b": 5.1, "capabilities": [ "coding", "chat", "general" ], "gguf_repo": "unsloth/gpt-oss-120b-GGUF", "mmproj": false, "quants": [ { "name": "Q4_K_M", "size_mb": 62769, "bpw": 4.3 }, { "name": "Q5_K_M", "size_mb": 62890, "bpw": 4.31 }, { "name": "UD-Q4_K_XL", "size_mb": 63016, "bpw": 4.32 }, { "name": "Q6_K", "size_mb": 63284, "bpw": 4.33 }, { "name": "UD-Q6_K_XL", "size_mb": 63284, "bpw": 4.33 }, { "name": "Q8_0", "size_mb": 63387, "bpw": 4.34 }, { "name": "UD-Q8_K_XL", "size_mb": 64473, "bpw": 4.41 } ], "notes": "MoE, 117B total, 5.1B active. Ships natively in MXFP4, so every GGUF quant is about the same size: take the native one.", "source": "https://huggingface.co/openai/gpt-oss-120b", "fetched": "2026-10-09T04:33+08:00" }, { "name": "gpt-oss-20b", "params_b": 20.9, "architecture": "moe", "active_params_b": 3.6, "capabilities": [ "coding", "chat", "general" ], "gguf_repo": "unsloth/gpt-oss-20b-GGUF", "mmproj": false, "quants": [ { "name": "Q4_K_M", "size_mb": 11625, "bpw": 4.45 }, { "name": "Q5_K_M", "size_mb": 11717, "bpw": 4.48 }, { "name": "UD-Q4_K_XL", "size_mb": 11872, "bpw": 4.54 }, { "name": "Q6_K", "size_mb": 12041, "bpw": 4.61 }, { "name": "UD-Q6_K_XL", "size_mb": 12041, "bpw": 4.61 }, { "name": "Q8_0", "size_mb": 12110, "bpw": 4.63 }, { "name": "UD-Q8_K_XL", "size_mb": 13195, "bpw": 5.05 } ], "notes": "MoE, 21B total, 3.6B active. Native MXFP4, fits a 16 GB GPU.", "source": "https://huggingface.co/openai/gpt-oss-20b", "fetched": "2026-10-09T04:33+08:00" }, { "name": "Qwen3-Coder-Next", "params_b": 79.7, "architecture": "moe", "active_params_b": 3, "capabilities": [ "coding", "general" ], "gguf_repo": "unsloth/Qwen3-Coder-Next-GGUF", "mmproj": false, "quants": [ { "name": "UD-Q2_K_XL", "size_mb": 26761, "bpw": 2.69 }, { "name": "UD-Q3_K_XL", "size_mb": 36283, "bpw": 3.64 }, { "name": "UD-IQ4_XS", "size_mb": 38429, "bpw": 3.86 }, { "name": "IQ4_XS", "size_mb": 42676, "bpw": 4.29 }, { "name": "MXFP4", "size_mb": 48034, "bpw": 4.82 }, { "name": "Q4_K_M", "size_mb": 48528, "bpw": 4.87 }, { "name": "UD-Q4_K_XL", "size_mb": 49608, "bpw": 4.98 }, { "name": "Q5_K_M", "size_mb": 56848, "bpw": 5.71 }, { "name": "UD-Q5_K_XL", "size_mb": 59541, "bpw": 5.98 }, { "name": "Q6_K", "size_mb": 65614, "bpw": 6.59 }, { "name": "UD-Q6_K_XL", "size_mb": 73119, "bpw": 7.34 }, { "name": "Q8_0", "size_mb": 84812, "bpw": 8.52 }, { "name": "UD-Q8_K_XL", "size_mb": 86349, "bpw": 8.67 } ], "notes": "MoE, 80B total, 3B active. Coding agent model.", "source": "https://huggingface.co/Qwen/Qwen3-Coder-Next", "fetched": "2026-10-09T04:33+08:00" }, { "name": "Qwen3-Coder-30B-A3B-Instruct", "params_b": 30.5, "architecture": "moe", "active_params_b": 3.3, "capabilities": [ "coding", "general" ], "gguf_repo": "unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF", "mmproj": false, "quants": [ { "name": "UD-Q2_K_XL", "size_mb": 11789, "bpw": 3.09 }, { "name": "UD-Q3_K_XL", "size_mb": 13806, "bpw": 3.62 }, { "name": "IQ4_XS", "size_mb": 16378, "bpw": 4.29 }, { "name": "UD-Q4_K_XL", "size_mb": 17665, "bpw": 4.63 }, { "name": "Q4_K_M", "size_mb": 18557, "bpw": 4.86 }, { "name": "Q5_K_M", "size_mb": 21726, "bpw": 5.69 }, { "name": "UD-Q5_K_XL", "size_mb": 21740, "bpw": 5.7 }, { "name": "Q6_K", "size_mb": 25093, "bpw": 6.57 }, { "name": "UD-Q6_K_XL", "size_mb": 26340, "bpw": 6.9 }, { "name": "Q8_0", "size_mb": 32484, "bpw": 8.51 }, { "name": "UD-Q8_K_XL", "size_mb": 35990, "bpw": 9.43 } ], "notes": "MoE, 30.5B total, 3.3B active. Very fast coder for 24 GB GPUs and unified-memory boxes.", "source": "https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct", "fetched": "2026-10-09T04:33+08:00" }, { "name": "Qwen3.6-35B-A3B", "params_b": 34.7, "architecture": "moe", "active_params_b": 3, "capabilities": [ "coding", "vision", "chat", "general" ], "gguf_repo": "unsloth/Qwen3.6-35B-A3B-GGUF", "mmproj": true, "quants": [ { "name": "UD-Q2_K_XL", "size_mb": 12291, "bpw": 2.84 }, { "name": "UD-Q3_K_XL", "size_mb": 16846, "bpw": 3.89 }, { "name": "UD-IQ4_XS", "size_mb": 17731, "bpw": 4.09 }, { "name": "MXFP4", "size_mb": 21706, "bpw": 5.01 }, { "name": "UD-Q4_K_XL", "size_mb": 22360, "bpw": 5.16 }, { "name": "UD-Q5_K_XL", "size_mb": 26593, "bpw": 6.14 }, { "name": "UD-Q6_K_XL", "size_mb": 31844, "bpw": 7.35 }, { "name": "Q8_0", "size_mb": 36903, "bpw": 8.52 }, { "name": "UD-Q8_K_XL", "size_mb": 38451, "bpw": 8.87 } ], "notes": "MoE, 35B total, ~3B active. MTP variant unsloth/Qwen3.6-35B-A3B-MTP-GGUF decodes faster.", "source": "https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF", "fetched": "2026-10-09T04:33+08:00" }, { "name": "Gemma-4-26B-A4B-it", "params_b": 25.2, "architecture": "moe", "active_params_b": 3.8, "capabilities": [ "vision", "chat", "general" ], "gguf_repo": "unsloth/gemma-4-26B-A4B-it-GGUF", "mmproj": true, "quants": [ { "name": "UD-Q2_K_XL", "size_mb": 10547, "bpw": 3.34 }, { "name": "UD-Q3_K_XL", "size_mb": 12907, "bpw": 4.09 }, { "name": "UD-IQ4_XS", "size_mb": 13597, "bpw": 4.31 }, { "name": "MXFP4", "size_mb": 16551, "bpw": 5.25 }, { "name": "UD-Q4_K_XL", "size_mb": 17011, "bpw": 5.39 }, { "name": "UD-Q5_K_XL", "size_mb": 21218, "bpw": 6.73 }, { "name": "UD-Q6_K_XL", "size_mb": 23295, "bpw": 7.39 }, { "name": "Q8_0", "size_mb": 26860, "bpw": 8.52 }, { "name": "UD-Q8_K_XL", "size_mb": 27636, "bpw": 8.76 } ], "notes": "MoE, 25.2B total, 3.8B active. Multimodal (mmproj). QAT build: unsloth/gemma-4-26B-A4B-it-qat-GGUF.", "source": "https://huggingface.co/google/gemma-4-26B-A4B-it", "fetched": "2026-10-09T04:33+08:00" }, { "name": "Nemotron-3-Nano-30B-A3B", "params_b": 31.6, "architecture": "moe", "active_params_b": 3.5, "capabilities": [ "chat", "general" ], "gguf_repo": "unsloth/Nemotron-3-Nano-30B-A3B-GGUF", "mmproj": false, "quants": [ { "name": "IQ4_XS", "size_mb": 18169, "bpw": 4.6 }, { "name": "UD-Q2_K_XL", "size_mb": 19920, "bpw": 5.05 }, { "name": "UD-Q3_K_XL", "size_mb": 19939, "bpw": 5.05 }, { "name": "UD-Q4_K_XL", "size_mb": 22834, "bpw": 5.78 }, { "name": "Q4_K_M", "size_mb": 24574, "bpw": 6.23 }, { "name": "Q5_K_M", "size_mb": 26149, "bpw": 6.62 }, { "name": "UD-Q5_K_XL", "size_mb": 27506, "bpw": 6.97 }, { "name": "Q6_K", "size_mb": 33508, "bpw": 8.49 }, { "name": "UD-Q6_K_XL", "size_mb": 33508, "bpw": 8.49 }, { "name": "Q8_0", "size_mb": 33585, "bpw": 8.51 }, { "name": "UD-Q8_K_XL", "size_mb": 40448, "bpw": 10.25 } ], "notes": "Hybrid Mamba MoE, 30B total, 3.5B active, 1M context.", "source": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16", "fetched": "2026-10-09T04:33+08:00" }, { "name": "Gemma-4-E4B-it", "params_b": 7.5, "architecture": "dense", "active_params_b": null, "capabilities": [ "vision", "chat", "general" ], "gguf_repo": "unsloth/gemma-4-E4B-it-GGUF", "mmproj": true, "quants": [ { "name": "UD-Q2_K_XL", "size_mb": 3757, "bpw": 4.0 }, { "name": "UD-Q3_K_XL", "size_mb": 4588, "bpw": 4.88 }, { "name": "IQ4_XS", "size_mb": 4715, "bpw": 5.02 }, { "name": "Q4_K_M", "size_mb": 4977, "bpw": 5.3 }, { "name": "UD-Q4_K_XL", "size_mb": 5126, "bpw": 5.45 }, { "name": "Q5_K_M", "size_mb": 5482, "bpw": 5.83 }, { "name": "UD-Q5_K_XL", "size_mb": 6656, "bpw": 7.08 }, { "name": "Q6_K", "size_mb": 7075, "bpw": 7.53 }, { "name": "UD-Q6_K_XL", "size_mb": 7458, "bpw": 7.94 }, { "name": "Q8_0", "size_mb": 8193, "bpw": 8.72 }, { "name": "UD-Q8_K_XL", "size_mb": 8712, "bpw": 9.27 } ], "notes": "Small multimodal model for laptops and 8 GB GPUs (mmproj for vision).", "source": "https://huggingface.co/google/gemma-4-E4B-it", "fetched": "2026-10-09T04:33+08:00" }, { "name": "Qwen3.5-9B", "params_b": 9.0, "architecture": "dense", "active_params_b": null, "capabilities": [ "chat", "general", "vision" ], "gguf_repo": "unsloth/Qwen3.5-9B-GGUF", "mmproj": true, "quants": [ { "name": "UD-Q2_K_XL", "size_mb": 4122, "bpw": 3.68 }, { "name": "UD-Q3_K_XL", "size_mb": 5054, "bpw": 4.52 }, { "name": "IQ4_XS", "size_mb": 5169, "bpw": 4.62 }, { "name": "Q4_K_M", "size_mb": 5681, "bpw": 5.08 }, { "name": "UD-Q4_K_XL", "size_mb": 5966, "bpw": 5.33 }, { "name": "Q5_K_M", "size_mb": 6578, "bpw": 5.88 }, { "name": "UD-Q5_K_XL", "size_mb": 6744, "bpw": 6.03 }, { "name": "Q6_K", "size_mb": 7458, "bpw": 6.66 }, { "name": "UD-Q6_K_XL", "size_mb": 8757, "bpw": 7.82 }, { "name": "Q8_0", "size_mb": 9528, "bpw": 8.51 }, { "name": "UD-Q8_K_XL", "size_mb": 12974, "bpw": 11.59 } ], "notes": "Small dense model for 8-12 GB GPUs.", "source": "https://huggingface.co/unsloth/Qwen3.5-9B-GGUF", "fetched": "2026-10-09T04:33+08:00" }, { "name": "Qwen3.5-4B", "params_b": 4.2, "architecture": "dense", "active_params_b": null, "capabilities": [ "chat", "general", "vision" ], "gguf_repo": "unsloth/Qwen3.5-4B-GGUF", "mmproj": true, "quants": [ { "name": "UD-Q2_K_XL", "size_mb": 1941, "bpw": 3.69 }, { "name": "UD-Q3_K_XL", "size_mb": 2436, "bpw": 4.63 }, { "name": "IQ4_XS", "size_mb": 2477, "bpw": 4.71 }, { "name": "Q4_K_M", "size_mb": 2741, "bpw": 5.21 }, { "name": "UD-Q4_K_XL", "size_mb": 2912, "bpw": 5.54 }, { "name": "Q5_K_M", "size_mb": 3144, "bpw": 5.98 }, { "name": "UD-Q5_K_XL", "size_mb": 3251, "bpw": 6.18 }, { "name": "Q6_K", "size_mb": 3526, "bpw": 6.71 }, { "name": "UD-Q6_K_XL", "size_mb": 4146, "bpw": 7.89 }, { "name": "Q8_0", "size_mb": 4482, "bpw": 8.53 }, { "name": "UD-Q8_K_XL", "size_mb": 5952, "bpw": 11.32 } ], "notes": "Tiny dense model for CPU-only boxes and 6 GB GPUs.", "source": "https://huggingface.co/unsloth/Qwen3.5-4B-GGUF", "fetched": "2026-10-09T04:33+08:00" } ], "hardware": [ { "id": "rtx-3090", "name": "NVIDIA GeForce RTX 3090", "vendor": "nvidia", "class": "consumer-gpu", "arch": "Ampere sm_86", "mem_gb": 24, "mem_type": "GDDR6X", "bw_gbps": 936, "fp4": false, "fp8": false, "engines": [ "llama.cpp", "exllamav3", "vllm", "ollama" ], "source": "https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3090-3090ti/", "confidence": "high", "match": [ "rtx 3090" ], "exclude": [ "3090 ti", "laptop", "mobile" ] }, { "id": "rtx-4090", "name": "NVIDIA GeForce RTX 4090", "vendor": "nvidia", "class": "consumer-gpu", "arch": "Ada sm_89", "mem_gb": 24, "mem_type": "GDDR6X", "bw_gbps": 1008, "fp4": false, "fp8": true, "engines": [ "llama.cpp", "exllamav3", "vllm", "sglang", "ollama" ], "source": "https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4090/", "confidence": "high", "match": [ "rtx 4090" ], "exclude": [ "laptop", "mobile", "4090 d" ] }, { "id": "rtx-5090", "name": "NVIDIA GeForce RTX 5090", "vendor": "nvidia", "class": "consumer-gpu", "arch": "Blackwell sm_120", "mem_gb": 32, "mem_type": "GDDR7", "bw_gbps": 1792, "fp4": true, "fp8": true, "engines": [ "llama.cpp", "exllamav3", "vllm", "sglang", "tensorrt-llm", "ollama" ], "source": "https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5090/", "confidence": "high", "match": [ "rtx 5090" ], "exclude": [ "laptop", "mobile" ] }, { "id": "rtx-pro-6000-bw", "name": "NVIDIA RTX PRO 6000 Blackwell", "vendor": "nvidia", "class": "workstation-gpu", "arch": "Blackwell sm_120", "mem_gb": 96, "mem_type": "GDDR7 ECC", "bw_gbps": 1792, "fp4": true, "fp8": true, "engines": [ "vllm", "sglang", "tensorrt-llm", "llama.cpp", "exllamav3" ], "source": "https://www.nvidia.com/en-us/products/workstations/professional-desktop-gpus/rtx-pro-6000/", "confidence": "high", "match": [ "rtx pro 6000" ], "exclude": [ "max-q" ] }, { "id": "dgx-spark", "name": "NVIDIA DGX Spark (GB10)", "vendor": "nvidia", "class": "unified-memory-box", "arch": "Blackwell GB10 sm_121, ARM64", "mem_gb": 128, "mem_type": "LPDDR5X unified", "bw_gbps": 273, "fp4": true, "fp8": true, "engines": [ "llama.cpp", "vllm", "sglang", "tensorrt-llm", "ollama" ], "source": "https://www.nvidia.com/en-us/products/workstations/dgx-spark/", "confidence": "high", "match": [ "gb10", "dgx spark" ], "exclude": [] }, { "id": "jetson-agx-thor", "name": "NVIDIA Jetson AGX Thor", "vendor": "nvidia", "class": "edge", "arch": "Blackwell, ARM64", "mem_gb": 128, "mem_type": "LPDDR5X unified", "bw_gbps": 273, "fp4": true, "fp8": true, "engines": [ "llama.cpp", "vllm", "tensorrt-llm" ], "source": "https://www.nvidia.com/en-us/autonomous-machines/embedded-systems/jetson-thor/", "confidence": "high", "match": [ "thor" ], "exclude": [] }, { "id": "h100-sxm", "name": "NVIDIA H100 SXM", "vendor": "nvidia", "class": "datacenter-gpu", "arch": "Hopper sm_90", "mem_gb": 80, "mem_type": "HBM3", "bw_gbps": 3350, "fp4": false, "fp8": true, "engines": [ "vllm", "sglang", "tensorrt-llm" ], "source": "https://www.nvidia.com/en-us/data-center/h100/", "confidence": "high", "match": [ "h100 80gb hbm3", "h100 sxm" ], "exclude": [ "pcie", "nvl" ] }, { "id": "h200", "name": "NVIDIA H200", "vendor": "nvidia", "class": "datacenter-gpu", "arch": "Hopper sm_90", "mem_gb": 141, "mem_type": "HBM3e", "bw_gbps": 4800, "fp4": false, "fp8": true, "engines": [ "vllm", "sglang", "tensorrt-llm" ], "source": "https://www.nvidia.com/en-us/data-center/h200/", "confidence": "high", "match": [ "h200" ], "exclude": [ "nvl" ] }, { "id": "b200", "name": "NVIDIA B200", "vendor": "nvidia", "class": "datacenter-gpu", "arch": "Blackwell sm_100", "mem_gb": 180, "mem_type": "HBM3e", "bw_gbps": 8000, "fp4": true, "fp8": true, "engines": [ "vllm", "sglang", "tensorrt-llm" ], "source": "https://www.nvidia.com/en-us/data-center/dgx-b200/", "confidence": "medium", "notes": "180 GB usable of 192 GB physical; vendors quote 7.7 to 8 TB/s.", "match": [ "b200" ], "exclude": [] }, { "id": "strix-halo-395", "name": "AMD Ryzen AI Max+ 395 (Strix Halo)", "vendor": "amd", "class": "unified-memory-box", "arch": "RDNA 3.5 gfx1151 iGPU + XDNA2 NPU", "mem_gb": 128, "mem_type": "LPDDR5X-8000 unified", "bw_gbps": 256, "fp4": false, "fp8": false, "engines": [ "llama.cpp", "lemonade", "ollama", "vllm" ], "notes": "llama.cpp Vulkan usually wins decode, ROCm wins prompt processing. vLLM only via community gfx1151 builds.", "source": "https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-plus-395.html", "confidence": "high", "match": [ "gfx1151", "8060s", "ai max+ 395", "ai max 395", "strix halo" ], "exclude": [] }, { "id": "rx-7900-xtx", "name": "AMD Radeon RX 7900 XTX", "vendor": "amd", "class": "consumer-gpu", "arch": "RDNA 3 gfx1100", "mem_gb": 24, "mem_type": "GDDR6", "bw_gbps": 960, "fp4": false, "fp8": false, "engines": [ "llama.cpp", "ollama", "vllm" ], "source": "https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xtx.html", "confidence": "high", "match": [ "7900 xtx" ], "exclude": [] }, { "id": "rx-9070-xt", "name": "AMD Radeon RX 9070 XT", "vendor": "amd", "class": "consumer-gpu", "arch": "RDNA 4 gfx1201", "mem_gb": 16, "mem_type": "GDDR6", "bw_gbps": 640, "fp4": false, "fp8": true, "engines": [ "llama.cpp", "ollama" ], "source": "https://www.amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9070xt.html", "confidence": "high", "match": [ "9070 xt" ], "exclude": [] }, { "id": "mi300x", "name": "AMD Instinct MI300X", "vendor": "amd", "class": "datacenter-gpu", "arch": "CDNA 3 gfx942", "mem_gb": 192, "mem_type": "HBM3", "bw_gbps": 5300, "fp4": false, "fp8": true, "engines": [ "vllm", "sglang" ], "source": "https://www.amd.com/en/products/accelerators/instinct/mi300/mi300x.html", "confidence": "high", "match": [ "mi300x" ], "exclude": [] }, { "id": "m4-max", "name": "Apple M4 Max (40-core GPU)", "vendor": "apple", "class": "unified-memory-box", "arch": "Apple Silicon, Metal", "mem_gb": 128, "mem_type": "LPDDR5X unified", "bw_gbps": 546, "fp4": false, "fp8": false, "engines": [ "mlx-lm", "llama.cpp", "ollama", "lm-studio" ], "notes": "Max memory depends on configuration (36 to 128 GB). mlx_lm.server is OpenAI-compatible, no wrapper needed.", "source": "https://www.apple.com/macbook-pro/specs/", "confidence": "high", "match": [ "m4 max" ], "exclude": [] }, { "id": "m3-ultra", "name": "Apple M3 Ultra", "vendor": "apple", "class": "unified-memory-box", "arch": "Apple Silicon, Metal", "mem_gb": 512, "mem_type": "LPDDR5 unified", "bw_gbps": 819, "fp4": false, "fp8": false, "engines": [ "mlx-lm", "llama.cpp", "ollama", "lm-studio" ], "notes": "Max memory depends on configuration (96 to 512 GB).", "source": "https://www.apple.com/mac-studio/specs/", "confidence": "high", "match": [ "m3 ultra" ], "exclude": [] }, { "id": "arc-b580", "name": "Intel Arc B580", "vendor": "intel", "class": "consumer-gpu", "arch": "Xe2 Battlemage", "mem_gb": 12, "mem_type": "GDDR6", "bw_gbps": 456, "fp4": false, "fp8": false, "engines": [ "llama.cpp", "openvino", "ollama" ], "notes": "llama.cpp via SYCL or Vulkan backend.", "source": "https://www.intel.com/content/www/us/en/products/sku/241598/intel-arc-b580-graphics/specifications.html", "confidence": "high", "match": [ "arc b580" ], "exclude": [] }, { "id": "arc-pro-b60", "name": "Intel Arc Pro B60", "vendor": "intel", "class": "workstation-gpu", "arch": "Xe2 Battlemage", "mem_gb": 24, "mem_type": "GDDR6", "bw_gbps": 456, "fp4": false, "fp8": false, "engines": [ "llama.cpp", "openvino", "vllm" ], "source": "https://download.intel.com/newsroom/2025/client-computing/Intel-Arc-Pro-B60-Data-Sheet.pdf", "confidence": "high", "match": [ "arc pro b60" ], "exclude": [] }, { "id": "rtx-3060-12gb", "name": "NVIDIA GeForce RTX 3060 12GB", "vendor": "nvidia", "class": "consumer-gpu", "arch": "Ampere sm_86", "mem_gb": 12, "mem_type": "GDDR6", "bw_gbps": 360, "fp4": false, "fp8": false, "engines": [ "llama.cpp", "exllamav3", "ollama", "vllm" ], "source": "https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3060-3060ti/", "confidence": "medium", "match": [ "rtx 3060" ], "exclude": [ "3060 ti", "laptop", "mobile" ], "notes": "The 8 GB RTX 3060 variant has 240 GB/s." }, { "id": "rtx-4060-ti", "name": "NVIDIA GeForce RTX 4060 Ti (16GB)", "vendor": "nvidia", "class": "consumer-gpu", "arch": "Ada sm_89", "mem_gb": 16, "mem_type": "GDDR6", "bw_gbps": 288, "fp4": false, "fp8": true, "engines": [ "llama.cpp", "exllamav3", "ollama", "vllm" ], "source": "https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4060-4060ti/", "confidence": "high", "match": [ "rtx 4060 ti" ], "exclude": [ "laptop", "mobile" ], "notes": "8 GB and 16 GB variants share the 288 GB/s bus." }, { "id": "rtx-4070-ti-super", "name": "NVIDIA GeForce RTX 4070 Ti SUPER", "vendor": "nvidia", "class": "consumer-gpu", "arch": "Ada sm_89", "mem_gb": 16, "mem_type": "GDDR6X", "bw_gbps": 672, "fp4": false, "fp8": true, "engines": [ "llama.cpp", "exllamav3", "ollama", "vllm" ], "source": "https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4070-family/", "confidence": "high", "match": [ "rtx 4070 ti super" ], "exclude": [ "laptop", "mobile" ] }, { "id": "rtx-4080-super", "name": "NVIDIA GeForce RTX 4080 SUPER", "vendor": "nvidia", "class": "consumer-gpu", "arch": "Ada sm_89", "mem_gb": 16, "mem_type": "GDDR6X", "bw_gbps": 736, "fp4": false, "fp8": true, "engines": [ "llama.cpp", "exllamav3", "ollama", "vllm" ], "source": "https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4080-family/", "confidence": "high", "match": [ "rtx 4080 super" ], "exclude": [ "laptop", "mobile" ] }, { "id": "rtx-4080", "name": "NVIDIA GeForce RTX 4080", "vendor": "nvidia", "class": "consumer-gpu", "arch": "Ada sm_89", "mem_gb": 16, "mem_type": "GDDR6X", "bw_gbps": 717, "fp4": false, "fp8": true, "engines": [ "llama.cpp", "exllamav3", "ollama", "vllm" ], "source": "https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4080-family/", "confidence": "high", "match": [ "rtx 4080" ], "exclude": [ "super", "laptop", "mobile" ] }, { "id": "rtx-5060-ti", "name": "NVIDIA GeForce RTX 5060 Ti (16GB)", "vendor": "nvidia", "class": "consumer-gpu", "arch": "Blackwell sm_120", "mem_gb": 16, "mem_type": "GDDR7", "bw_gbps": 448, "fp4": true, "fp8": true, "engines": [ "llama.cpp", "exllamav3", "ollama", "vllm" ], "source": "https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5060-family/", "confidence": "high", "match": [ "rtx 5060 ti" ], "exclude": [ "laptop", "mobile" ] }, { "id": "rtx-5070-ti", "name": "NVIDIA GeForce RTX 5070 Ti", "vendor": "nvidia", "class": "consumer-gpu", "arch": "Blackwell sm_120", "mem_gb": 16, "mem_type": "GDDR7", "bw_gbps": 896, "fp4": true, "fp8": true, "engines": [ "llama.cpp", "exllamav3", "ollama", "vllm" ], "source": "https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5070-family/", "confidence": "high", "match": [ "rtx 5070 ti" ], "exclude": [ "laptop", "mobile" ] }, { "id": "rtx-5080", "name": "NVIDIA GeForce RTX 5080", "vendor": "nvidia", "class": "consumer-gpu", "arch": "Blackwell sm_120", "mem_gb": 16, "mem_type": "GDDR7", "bw_gbps": 960, "fp4": true, "fp8": true, "engines": [ "llama.cpp", "exllamav3", "ollama", "vllm" ], "source": "https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5080/", "confidence": "high", "match": [ "rtx 5080" ], "exclude": [ "laptop", "mobile" ] }, { "id": "rtx-6000-ada", "name": "NVIDIA RTX 6000 Ada Generation", "vendor": "nvidia", "class": "workstation-gpu", "arch": "Ada sm_89", "mem_gb": 48, "mem_type": "GDDR6 ECC", "bw_gbps": 960, "fp4": false, "fp8": true, "engines": [ "vllm", "sglang", "llama.cpp", "exllamav3" ], "source": "https://www.nvidia.com/en-us/products/workstations/rtx-6000/", "confidence": "high", "match": [ "rtx 6000 ada" ], "exclude": [] }, { "id": "rtx-a6000", "name": "NVIDIA RTX A6000", "vendor": "nvidia", "class": "workstation-gpu", "arch": "Ampere sm_86", "mem_gb": 48, "mem_type": "GDDR6 ECC", "bw_gbps": 768, "fp4": false, "fp8": false, "engines": [ "vllm", "llama.cpp", "exllamav3", "sglang" ], "source": "https://www.nvidia.com/en-us/products/workstations/rtx-a6000/", "confidence": "high", "match": [ "rtx a6000" ], "exclude": [ "ada" ] }, { "id": "l40s", "name": "NVIDIA L40S", "vendor": "nvidia", "class": "datacenter-gpu", "arch": "Ada sm_89", "mem_gb": 48, "mem_type": "GDDR6 ECC", "bw_gbps": 864, "fp4": false, "fp8": true, "engines": [ "vllm", "sglang", "tensorrt-llm", "llama.cpp" ], "source": "https://www.nvidia.com/en-us/data-center/l40s/", "confidence": "high", "match": [ "l40s" ], "exclude": [] }, { "id": "a100-80gb", "name": "NVIDIA A100 80GB", "vendor": "nvidia", "class": "datacenter-gpu", "arch": "Ampere sm_80", "mem_gb": 80, "mem_type": "HBM2e", "bw_gbps": 1935, "fp4": false, "fp8": false, "engines": [ "vllm", "sglang", "tensorrt-llm" ], "source": "https://www.nvidia.com/en-us/data-center/a100/", "confidence": "high", "match": [ "a100 80gb", "a100 sxm4 80gb" ], "exclude": [], "notes": "PCIe 1935 GB/s; the SXM part reaches 2039 GB/s." }, { "id": "m4-pro", "name": "Apple M4 Pro", "vendor": "apple", "class": "unified-memory-box", "arch": "Apple Silicon, Metal", "mem_gb": 64, "mem_type": "LPDDR5X unified", "bw_gbps": 273, "fp4": false, "fp8": false, "engines": [ "mlx-lm", "llama.cpp", "ollama", "lm-studio" ], "source": "https://www.apple.com/mac-mini/specs/", "confidence": "high", "match": [ "m4 pro" ], "exclude": [] }, { "id": "m3-max", "name": "Apple M3 Max", "vendor": "apple", "class": "unified-memory-box", "arch": "Apple Silicon, Metal", "mem_gb": 128, "mem_type": "LPDDR5 unified", "bw_gbps": 300, "fp4": false, "fp8": false, "engines": [ "mlx-lm", "llama.cpp", "ollama", "lm-studio" ], "source": "https://support.apple.com/specs", "confidence": "medium", "match": [ "m3 max" ], "exclude": [], "notes": "300 GB/s on the 30-core GPU part, 400 GB/s on the 40-core part; we take the lower." }, { "id": "m2-max", "name": "Apple M2 Max", "vendor": "apple", "class": "unified-memory-box", "arch": "Apple Silicon, Metal", "mem_gb": 96, "mem_type": "LPDDR5 unified", "bw_gbps": 400, "fp4": false, "fp8": false, "engines": [ "mlx-lm", "llama.cpp", "ollama", "lm-studio" ], "source": "https://support.apple.com/specs", "confidence": "high", "match": [ "m2 max" ], "exclude": [] }, { "id": "m1-max", "name": "Apple M1 Max", "vendor": "apple", "class": "unified-memory-box", "arch": "Apple Silicon, Metal", "mem_gb": 64, "mem_type": "LPDDR5 unified", "bw_gbps": 400, "fp4": false, "fp8": false, "engines": [ "mlx-lm", "llama.cpp", "ollama", "lm-studio" ], "source": "https://support.apple.com/specs", "confidence": "high", "match": [ "m1 max" ], "exclude": [] }, { "id": "m2-ultra", "name": "Apple M2 Ultra", "vendor": "apple", "class": "unified-memory-box", "arch": "Apple Silicon, Metal", "mem_gb": 192, "mem_type": "LPDDR5 unified", "bw_gbps": 800, "fp4": false, "fp8": false, "engines": [ "mlx-lm", "llama.cpp", "ollama", "lm-studio" ], "source": "https://support.apple.com/specs", "confidence": "high", "match": [ "m2 ultra" ], "exclude": [] }, { "id": "m1-ultra", "name": "Apple M1 Ultra", "vendor": "apple", "class": "unified-memory-box", "arch": "Apple Silicon, Metal", "mem_gb": 128, "mem_type": "LPDDR5 unified", "bw_gbps": 800, "fp4": false, "fp8": false, "engines": [ "mlx-lm", "llama.cpp", "ollama", "lm-studio" ], "source": "https://support.apple.com/specs", "confidence": "high", "match": [ "m1 ultra" ], "exclude": [] } ], "matrix": { "rtx-3090": { "general": [ { "model": "Qwen3.6-35B-A3B", "quant": "MXFP4", "size_gb": 21.7, "est_tps": [ 117, 205 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q5_K_XL", "size_gb": 21.7, "est_tps": [ 102, 179 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q5_K_XL", "size_gb": 20.9, "est_tps": [ 24, 33 ] } ], "coding": [ { "model": "Qwen3.6-35B-A3B", "quant": "MXFP4", "size_gb": 21.7, "est_tps": [ 117, 205 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q5_K_XL", "size_gb": 21.7, "est_tps": [ 102, 179 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q5_K_XL", "size_gb": 20.9, "est_tps": [ 24, 33 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "MXFP4", "size_gb": 21.7, "est_tps": [ 117, 205 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q5_K_XL", "size_gb": 20.9, "est_tps": [ 24, 33 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q5_K_XL", "size_gb": 21.2, "est_tps": [ 83, 145 ] } ] }, "rtx-4090": { "general": [ { "model": "Qwen3.6-35B-A3B", "quant": "MXFP4", "size_gb": 21.7, "est_tps": [ 123, 214 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q5_K_XL", "size_gb": 21.7, "est_tps": [ 107, 188 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q5_K_XL", "size_gb": 20.9, "est_tps": [ 26, 36 ] } ], "coding": [ { "model": "Qwen3.6-35B-A3B", "quant": "MXFP4", "size_gb": 21.7, "est_tps": [ 123, 214 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q5_K_XL", "size_gb": 21.7, "est_tps": [ 107, 188 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q5_K_XL", "size_gb": 20.9, "est_tps": [ 26, 36 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "MXFP4", "size_gb": 21.7, "est_tps": [ 123, 214 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q5_K_XL", "size_gb": 20.9, "est_tps": [ 26, 36 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q5_K_XL", "size_gb": 21.2, "est_tps": [ 88, 153 ] } ] }, "rtx-5090": { "general": [ { "model": "Qwen3-Coder-Next", "quant": "UD-Q2_K_XL", "size_gb": 26.8, "est_tps": [ 204, 357 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q5_K_XL", "size_gb": 26.6, "est_tps": [ 149, 261 ] }, { "model": "Nemotron-3-Nano-30B-A3B", "quant": "UD-Q5_K_XL", "size_gb": 27.5, "est_tps": [ 129, 226 ] } ], "coding": [ { "model": "Qwen3-Coder-Next", "quant": "UD-Q2_K_XL", "size_gb": 26.8, "est_tps": [ 204, 357 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q5_K_XL", "size_gb": 26.6, "est_tps": [ 149, 261 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q6_K_XL", "size_gb": 26.3, "est_tps": [ 134, 234 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q5_K_XL", "size_gb": 26.6, "est_tps": [ 149, 261 ] }, { "model": "Qwen3.8-27B", "quant": "Q8_0", "size_gb": 29.0, "est_tps": [ 33, 45 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q8_K_XL", "size_gb": 27.6, "est_tps": [ 107, 188 ] } ] }, "rtx-pro-6000-bw": { "general": [ { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 135, 236 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q2_K_XL", "size_gb": 78.9, "est_tps": [ 138, 242 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q8_K_XL", "size_gb": 86.3, "est_tps": [ 124, 218 ] } ], "coding": [ { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 135, 236 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q2_K_XL", "size_gb": 78.9, "est_tps": [ 138, 242 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q8_K_XL", "size_gb": 86.3, "est_tps": [ 124, 218 ] } ], "vision": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q2_K_XL", "size_gb": 78.9, "est_tps": [ 138, 242 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 123, 215 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 30, 42 ] } ] }, "dgx-spark": { "general": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 31, 54 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 34, 60 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q8_K_XL", "size_gb": 86.3, "est_tps": [ 30, 53 ] } ], "coding": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 31, 54 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 34, 60 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q8_K_XL", "size_gb": 86.3, "est_tps": [ 30, 53 ] } ], "vision": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 31, 54 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 29, 52 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 4.8, 6.5 ] } ] }, "jetson-agx-thor": { "general": [ { "model": "GLM-5.3-Flash", "quant": "UD-Q2_K_XL", "size_gb": 108.7, "est_tps": [ 17, 29 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q4_K_XL", "size_gb": 111.3, "est_tps": [ 26, 46 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 34, 60 ] } ], "coding": [ { "model": "GLM-5.3-Flash", "quant": "UD-Q2_K_XL", "size_gb": 108.7, "est_tps": [ 17, 29 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q4_K_XL", "size_gb": 111.3, "est_tps": [ 26, 46 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 34, 60 ] } ], "vision": [ { "model": "GLM-5.3-Flash", "quant": "UD-Q2_K_XL", "size_gb": 108.7, "est_tps": [ 17, 29 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q4_K_XL", "size_gb": 111.3, "est_tps": [ 26, 46 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 29, 52 ] } ] }, "h100-sxm": { "general": [ { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 179, 313 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q6_K_XL", "size_gb": 73.1, "est_tps": [ 180, 315 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 167, 293 ] } ], "coding": [ { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 179, 313 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q6_K_XL", "size_gb": 73.1, "est_tps": [ 180, 315 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 167, 293 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 167, 293 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 55, 77 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q8_K_XL", "size_gb": 27.6, "est_tps": [ 151, 265 ] } ] }, "h200": { "general": [ { "model": "GLM-5.3-Flash", "quant": "UD-Q2_K_XL", "size_gb": 108.7, "est_tps": [ 150, 262 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q4_K_XL", "size_gb": 111.3, "est_tps": [ 183, 320 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 201, 353 ] } ], "coding": [ { "model": "GLM-5.3-Flash", "quant": "UD-Q2_K_XL", "size_gb": 108.7, "est_tps": [ 150, 262 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q4_K_XL", "size_gb": 111.3, "est_tps": [ 183, 320 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 201, 353 ] } ], "vision": [ { "model": "GLM-5.3-Flash", "quant": "UD-Q2_K_XL", "size_gb": 108.7, "est_tps": [ 150, 262 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q4_K_XL", "size_gb": 111.3, "est_tps": [ 183, 320 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 191, 334 ] } ] }, "b200": { "general": [ { "model": "GLM-5.3-Flash", "quant": "UD-IQ4_XS", "size_gb": 156.8, "est_tps": [ 160, 280 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q5_K_XL", "size_gb": 158.3, "est_tps": [ 193, 338 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 228, 400 ] } ], "coding": [ { "model": "GLM-5.3-Flash", "quant": "UD-IQ4_XS", "size_gb": 156.8, "est_tps": [ 160, 280 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q5_K_XL", "size_gb": 158.3, "est_tps": [ 193, 338 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 228, 400 ] } ], "vision": [ { "model": "GLM-5.3-Flash", "quant": "UD-IQ4_XS", "size_gb": 156.8, "est_tps": [ 160, 280 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q5_K_XL", "size_gb": 158.3, "est_tps": [ 193, 338 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 220, 386 ] } ] }, "strix-halo-395": { "general": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 29, 51 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 32, 57 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q8_K_XL", "size_gb": 86.3, "est_tps": [ 28, 50 ] } ], "coding": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 29, 51 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 32, 57 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q8_K_XL", "size_gb": 86.3, "est_tps": [ 28, 50 ] } ], "vision": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 29, 51 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 28, 49 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 4.5, 6.1 ] } ] }, "rx-7900-xtx": { "general": [ { "model": "Qwen3.6-35B-A3B", "quant": "MXFP4", "size_gb": 21.7, "est_tps": [ 119, 209 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q5_K_XL", "size_gb": 21.7, "est_tps": [ 104, 182 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q5_K_XL", "size_gb": 20.9, "est_tps": [ 25, 34 ] } ], "coding": [ { "model": "Qwen3.6-35B-A3B", "quant": "MXFP4", "size_gb": 21.7, "est_tps": [ 119, 209 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q5_K_XL", "size_gb": 21.7, "est_tps": [ 104, 182 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q5_K_XL", "size_gb": 20.9, "est_tps": [ 25, 34 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "MXFP4", "size_gb": 21.7, "est_tps": [ 119, 209 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q5_K_XL", "size_gb": 20.9, "est_tps": [ 25, 34 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q5_K_XL", "size_gb": 21.2, "est_tps": [ 85, 148 ] } ] }, "rx-9070-xt": { "general": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 24, 33 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 107, 188 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 87, 152 ] } ], "coding": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 24, 33 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 107, 188 ] }, { "model": "gpt-oss-20b", "quant": "UD-Q8_K_XL", "size_gb": 13.2, "est_tps": [ 81, 141 ] } ], "vision": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 24, 33 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 87, 152 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q2_K_XL", "size_gb": 12.3, "est_tps": [ 131, 228 ] } ] }, "mi300x": { "general": [ { "model": "GLM-5.3-Flash", "quant": "UD-IQ4_XS", "size_gb": 156.8, "est_tps": [ 131, 229 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q6_K_XL", "size_gb": 169.2, "est_tps": [ 161, 282 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 207, 363 ] } ], "coding": [ { "model": "GLM-5.3-Flash", "quant": "UD-IQ4_XS", "size_gb": 156.8, "est_tps": [ 131, 229 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q6_K_XL", "size_gb": 169.2, "est_tps": [ 161, 282 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 207, 363 ] } ], "vision": [ { "model": "GLM-5.3-Flash", "quant": "UD-IQ4_XS", "size_gb": 156.8, "est_tps": [ 131, 229 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q6_K_XL", "size_gb": 169.2, "est_tps": [ 161, 282 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 197, 345 ] } ] }, "m4-max": { "general": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 55, 97 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 61, 107 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q8_K_XL", "size_gb": 86.3, "est_tps": [ 54, 95 ] } ], "coding": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 55, 97 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 61, 107 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q8_K_XL", "size_gb": 86.3, "est_tps": [ 54, 95 ] } ], "vision": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 55, 97 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 53, 93 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 9.5, 13 ] } ] }, "m3-ultra": { "general": [ { "model": "GLM-5.3-Flash", "quant": "Q8_0", "size_gb": 341.0, "est_tps": [ 16, 28 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "Q8_0", "size_gb": 188.2, "est_tps": [ 44, 76 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 83, 145 ] } ], "coding": [ { "model": "GLM-5.3-Flash", "quant": "Q8_0", "size_gb": 341.0, "est_tps": [ 16, 28 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "Q8_0", "size_gb": 188.2, "est_tps": [ 44, 76 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 83, 145 ] } ], "vision": [ { "model": "GLM-5.3-Flash", "quant": "Q8_0", "size_gb": 341.0, "est_tps": [ 16, 28 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "Q8_0", "size_gb": 188.2, "est_tps": [ 44, 76 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 73, 128 ] } ] }, "arc-b580": { "general": [ { "model": "Qwen3.8-27B", "quant": "UD-Q2_K_XL", "size_gb": 9.8, "est_tps": [ 25, 34 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q2_K_XL", "size_gb": 10.5, "est_tps": [ 82, 143 ] }, { "model": "Qwen3.5-9B", "quant": "Q8_0", "size_gb": 9.5, "est_tps": [ 26, 35 ] } ], "coding": [ { "model": "Qwen3.8-27B", "quant": "UD-Q2_K_XL", "size_gb": 9.8, "est_tps": [ 25, 34 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q2_K_XL", "size_gb": 10.5, "est_tps": [ 82, 143 ] }, { "model": "Qwen3.5-9B", "quant": "Q8_0", "size_gb": 9.5, "est_tps": [ 26, 35 ] } ], "vision": [ { "model": "Qwen3.8-27B", "quant": "UD-Q2_K_XL", "size_gb": 9.8, "est_tps": [ 25, 34 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q2_K_XL", "size_gb": 10.5, "est_tps": [ 82, 143 ] }, { "model": "Qwen3.5-9B", "quant": "Q8_0", "size_gb": 9.5, "est_tps": [ 26, 35 ] } ] }, "arc-pro-b60": { "general": [ { "model": "Qwen3.6-35B-A3B", "quant": "MXFP4", "size_gb": 21.7, "est_tps": [ 72, 127 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q5_K_XL", "size_gb": 21.7, "est_tps": [ 61, 107 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q5_K_XL", "size_gb": 20.9, "est_tps": [ 12, 16 ] } ], "coding": [ { "model": "Qwen3.6-35B-A3B", "quant": "MXFP4", "size_gb": 21.7, "est_tps": [ 72, 127 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q5_K_XL", "size_gb": 21.7, "est_tps": [ 61, 107 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q5_K_XL", "size_gb": 20.9, "est_tps": [ 12, 16 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "MXFP4", "size_gb": 21.7, "est_tps": [ 72, 127 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q5_K_XL", "size_gb": 20.9, "est_tps": [ 12, 16 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q5_K_XL", "size_gb": 21.2, "est_tps": [ 48, 83 ] } ] }, "rtx-3060-12gb": { "general": [ { "model": "Qwen3.8-27B", "quant": "UD-Q2_K_XL", "size_gb": 9.8, "est_tps": [ 20, 27 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q2_K_XL", "size_gb": 10.5, "est_tps": [ 69, 121 ] }, { "model": "Qwen3.5-9B", "quant": "Q8_0", "size_gb": 9.5, "est_tps": [ 20, 28 ] } ], "coding": [ { "model": "Qwen3.8-27B", "quant": "UD-Q2_K_XL", "size_gb": 9.8, "est_tps": [ 20, 27 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q2_K_XL", "size_gb": 10.5, "est_tps": [ 69, 121 ] }, { "model": "Qwen3.5-9B", "quant": "Q8_0", "size_gb": 9.5, "est_tps": [ 20, 28 ] } ], "vision": [ { "model": "Qwen3.8-27B", "quant": "UD-Q2_K_XL", "size_gb": 9.8, "est_tps": [ 20, 27 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q2_K_XL", "size_gb": 10.5, "est_tps": [ 69, 121 ] }, { "model": "Qwen3.5-9B", "quant": "Q8_0", "size_gb": 9.5, "est_tps": [ 20, 28 ] } ] }, "rtx-4060-ti": { "general": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 11, 15 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 61, 106 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 47, 82 ] } ], "coding": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 11, 15 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 61, 106 ] }, { "model": "gpt-oss-20b", "quant": "UD-Q8_K_XL", "size_gb": 13.2, "est_tps": [ 43, 75 ] } ], "vision": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 11, 15 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 47, 82 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q2_K_XL", "size_gb": 12.3, "est_tps": [ 78, 137 ] } ] }, "rtx-4070-ti-super": { "general": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 25, 35 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 110, 193 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 90, 157 ] } ], "coding": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 25, 35 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 110, 193 ] }, { "model": "gpt-oss-20b", "quant": "UD-Q8_K_XL", "size_gb": 13.2, "est_tps": [ 84, 146 ] } ], "vision": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 25, 35 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 90, 157 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q2_K_XL", "size_gb": 12.3, "est_tps": [ 134, 235 ] } ] }, "rtx-4080-super": { "general": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 28, 38 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 117, 204 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 96, 167 ] } ], "coding": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 28, 38 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 117, 204 ] }, { "model": "gpt-oss-20b", "quant": "UD-Q8_K_XL", "size_gb": 13.2, "est_tps": [ 89, 156 ] } ], "vision": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 28, 38 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 96, 167 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q2_K_XL", "size_gb": 12.3, "est_tps": [ 140, 246 ] } ] }, "rtx-4080": { "general": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 27, 37 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 115, 201 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 94, 165 ] } ], "coding": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 27, 37 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 115, 201 ] }, { "model": "gpt-oss-20b", "quant": "UD-Q8_K_XL", "size_gb": 13.2, "est_tps": [ 88, 153 ] } ], "vision": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 27, 37 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 94, 165 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q2_K_XL", "size_gb": 12.3, "est_tps": [ 139, 243 ] } ] }, "rtx-5060-ti": { "general": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 17, 23 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 85, 148 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 67, 117 ] } ], "coding": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 17, 23 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 85, 148 ] }, { "model": "gpt-oss-20b", "quant": "UD-Q8_K_XL", "size_gb": 13.2, "est_tps": [ 62, 108 ] } ], "vision": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 17, 23 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 67, 117 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q2_K_XL", "size_gb": 12.3, "est_tps": [ 106, 185 ] } ] }, "rtx-5070-ti": { "general": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 33, 46 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 130, 228 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 109, 190 ] } ], "coding": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 33, 46 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 130, 228 ] }, { "model": "gpt-oss-20b", "quant": "UD-Q8_K_XL", "size_gb": 13.2, "est_tps": [ 102, 178 ] } ], "vision": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 33, 46 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 109, 190 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q2_K_XL", "size_gb": 12.3, "est_tps": [ 155, 270 ] } ] }, "rtx-5080": { "general": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 36, 49 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 135, 237 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 113, 198 ] } ], "coding": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 36, 49 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q3_K_XL", "size_gb": 13.8, "est_tps": [ 135, 237 ] }, { "model": "gpt-oss-20b", "quant": "UD-Q8_K_XL", "size_gb": 13.2, "est_tps": [ 106, 186 ] } ], "vision": [ { "model": "Qwen3.8-27B", "quant": "UD-IQ4_XS", "size_gb": 14.3, "est_tps": [ 36, 49 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-IQ4_XS", "size_gb": 13.6, "est_tps": [ 113, 198 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q2_K_XL", "size_gb": 12.3, "est_tps": [ 159, 279 ] } ] }, "rtx-6000-ada": { "general": [ { "model": "Qwen3-Coder-Next", "quant": "IQ4_XS", "size_gb": 42.7, "est_tps": [ 130, 228 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 82, 144 ] }, { "model": "Nemotron-3-Nano-30B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 40.4, "est_tps": [ 66, 115 ] } ], "coding": [ { "model": "Qwen3-Coder-Next", "quant": "IQ4_XS", "size_gb": 42.7, "est_tps": [ 130, 228 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 82, 144 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q8_K_XL", "size_gb": 36.0, "est_tps": [ 73, 128 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 82, 144 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 17, 23 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q8_K_XL", "size_gb": 27.6, "est_tps": [ 70, 122 ] } ] }, "rtx-a6000": { "general": [ { "model": "Qwen3-Coder-Next", "quant": "IQ4_XS", "size_gb": 42.7, "est_tps": [ 114, 200 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 70, 122 ] }, { "model": "Nemotron-3-Nano-30B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 40.4, "est_tps": [ 55, 97 ] } ], "coding": [ { "model": "Qwen3-Coder-Next", "quant": "IQ4_XS", "size_gb": 42.7, "est_tps": [ 114, 200 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 70, 122 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q8_K_XL", "size_gb": 36.0, "est_tps": [ 62, 108 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 70, 122 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 13, 18 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q8_K_XL", "size_gb": 27.6, "est_tps": [ 59, 103 ] } ] }, "l40s": { "general": [ { "model": "Qwen3-Coder-Next", "quant": "IQ4_XS", "size_gb": 42.7, "est_tps": [ 123, 215 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 76, 133 ] }, { "model": "Nemotron-3-Nano-30B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 40.4, "est_tps": [ 61, 106 ] } ], "coding": [ { "model": "Qwen3-Coder-Next", "quant": "IQ4_XS", "size_gb": 42.7, "est_tps": [ 123, 215 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 76, 133 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q8_K_XL", "size_gb": 36.0, "est_tps": [ 68, 119 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 76, 133 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 15, 20 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q8_K_XL", "size_gb": 27.6, "est_tps": [ 64, 113 ] } ] }, "a100-80gb": { "general": [ { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 140, 245 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q6_K_XL", "size_gb": 73.1, "est_tps": [ 142, 248 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 128, 224 ] } ], "coding": [ { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 140, 245 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q6_K_XL", "size_gb": 73.1, "est_tps": [ 142, 248 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 128, 224 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 128, 224 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 33, 45 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q8_K_XL", "size_gb": 27.6, "est_tps": [ 113, 197 ] } ] }, "m4-pro": { "general": [ { "model": "Qwen3-Coder-Next", "quant": "UD-Q4_K_XL", "size_gb": 49.6, "est_tps": [ 49, 85 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 29, 52 ] }, { "model": "Nemotron-3-Nano-30B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 40.4, "est_tps": [ 22, 39 ] } ], "coding": [ { "model": "Qwen3-Coder-Next", "quant": "UD-Q4_K_XL", "size_gb": 49.6, "est_tps": [ 49, 85 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 29, 52 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q8_K_XL", "size_gb": 36.0, "est_tps": [ 26, 45 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 29, 52 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 4.8, 6.5 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q8_K_XL", "size_gb": 27.6, "est_tps": [ 24, 42 ] } ] }, "m3-max": { "general": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 33, 58 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 37, 65 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q8_K_XL", "size_gb": 86.3, "est_tps": [ 33, 57 ] } ], "coding": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 33, 58 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 37, 65 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q8_K_XL", "size_gb": 86.3, "est_tps": [ 33, 57 ] } ], "vision": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 33, 58 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 32, 56 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 5.2, 7.1 ] } ] }, "m2-max": { "general": [ { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 47, 83 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q6_K_XL", "size_gb": 73.1, "est_tps": [ 48, 85 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 41, 72 ] } ], "coding": [ { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 47, 83 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q6_K_XL", "size_gb": 73.1, "est_tps": [ 48, 85 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 41, 72 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 41, 72 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 7.0, 9.5 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q8_K_XL", "size_gb": 27.6, "est_tps": [ 34, 59 ] } ] }, "m1-max": { "general": [ { "model": "Qwen3-Coder-Next", "quant": "UD-Q4_K_XL", "size_gb": 49.6, "est_tps": [ 66, 115 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 41, 72 ] }, { "model": "Nemotron-3-Nano-30B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 40.4, "est_tps": [ 32, 56 ] } ], "coding": [ { "model": "Qwen3-Coder-Next", "quant": "UD-Q4_K_XL", "size_gb": 49.6, "est_tps": [ 66, 115 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 41, 72 ] }, { "model": "Qwen3-Coder-30B-A3B-Instruct", "quant": "UD-Q8_K_XL", "size_gb": 36.0, "est_tps": [ 36, 63 ] } ], "vision": [ { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 41, 72 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 7.0, 9.5 ] }, { "model": "Gemma-4-26B-A4B-it", "quant": "UD-Q8_K_XL", "size_gb": 27.6, "est_tps": [ 34, 59 ] } ] }, "m2-ultra": { "general": [ { "model": "GLM-5.3-Flash", "quant": "UD-Q3_K_XL", "size_gb": 147.5, "est_tps": [ 34, 60 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q4_K_XL", "size_gb": 111.3, "est_tps": [ 65, 114 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 81, 142 ] } ], "coding": [ { "model": "GLM-5.3-Flash", "quant": "UD-Q3_K_XL", "size_gb": 147.5, "est_tps": [ 34, 60 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q4_K_XL", "size_gb": 111.3, "est_tps": [ 65, 114 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 81, 142 ] } ], "vision": [ { "model": "GLM-5.3-Flash", "quant": "UD-Q3_K_XL", "size_gb": 147.5, "est_tps": [ 34, 60 ] }, { "model": "Qwen3.8-Flash-Next", "quant": "UD-Q4_K_XL", "size_gb": 111.3, "est_tps": [ 65, 114 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 72, 126 ] } ] }, "m1-ultra": { "general": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 74, 130 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 81, 142 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q8_K_XL", "size_gb": 86.3, "est_tps": [ 73, 128 ] } ], "coding": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 74, 130 ] }, { "model": "gpt-oss-120b", "quant": "UD-Q8_K_XL", "size_gb": 64.5, "est_tps": [ 81, 142 ] }, { "model": "Qwen3-Coder-Next", "quant": "UD-Q8_K_XL", "size_gb": 86.3, "est_tps": [ 73, 128 ] } ], "vision": [ { "model": "Qwen3.8-Flash-Next", "quant": "UD-IQ4_XS", "size_gb": 93.7, "est_tps": [ 74, 130 ] }, { "model": "Qwen3.6-35B-A3B", "quant": "UD-Q8_K_XL", "size_gb": 38.5, "est_tps": [ 72, 126 ] }, { "model": "Qwen3.8-27B", "quant": "UD-Q8_K_XL", "size_gb": 31.5, "est_tps": [ 14, 19 ] } ] } }, "unknown_hardware": [ "Apple M-series Max (MacBook Pro / Mac Studio Max)", "Apple Mac Studio (M-series Ultra)" ], "prices": { "generated": "2026-10-09T04:33+08:00", "currency": "SGD", "gst_rate": 0.09, "fx": { "usd_sgd": 1.282, "date": "2026-10-08", "source": "https://api.frankfurter.dev/v1/latest?base=USD&symbols=SGD" }, "items": [ { "id": "strix-halo-128", "device": "strix-halo-395", "label": "AMD Ryzen AI Max+ 395 mini PC, 128 GB", "class": "small", "from_sgd": 4324.74, "to_sgd": 5093.66, "sources_ok": 0, "sources": [ { "seller": "GMKtec (EVO-X2)", "trust": "Official store", "url": "https://www.gmktec.com/products/amd-ryzen%e2%84%a2-ai-max-395-evo-x2-ai-mini-pc", "ships_from": "overseas", "checked": "2026-10-09T04:33+08:00", "ok": false, "error": "HTTPError: HTTP Error 429: Too Many Requests" }, { "seller": "Bosgame (M5)", "trust": "Official store", "url": "https://www.bosgame.com/products/bosgame-m5-ai-mini-desktop-ryzen-ai-max-395-96gb-128gb-2tb", "ships_from": "overseas", "checked": "2026-10-09T04:33+08:00", "ok": false, "error": "HTTPError: HTTP Error 429: Too Many Requests" }, { "seller": "ALLSTARS (Singapore)", "trust": "Local distributor, 3yr warranty", "url": "https://allstars.com.sg/shop/pc/pc-complete/gmktec-evo-x2-ai-mini-pc-amd-ryzen-ai-max-395-128gb-lpddr5-2tb-gen4-ssd-2-5g-lan-wifi7-licensed-windows-11-pro/", "ships_from": "sg", "checked": "2026-10-09T04:33+08:00", "ok": false, "out_of_stock": true, "listed": 5599.0, "currency": "SGD", "error": "ValueError: out of stock (page still lists SGD 5,599)" }, { "seller": "Framework (Desktop)", "trust": "Official store, configure to order", "url": "https://frame.work/sg/en/desktop", "ships_from": "overseas", "checked": "2026-10-09T04:33+08:00", "ok": false, "link_only": true } ], "stale": true, "last_read": "2026-10-06T04:34+08:00" }, { "id": "gb10-128", "device": "dgx-spark", "label": "NVIDIA GB10 box, 128 GB (DGX Spark or ASUS Ascent GX10)", "class": "small", "from_sgd": 6624.0, "to_sgd": 8096.0, "sources_ok": 0, "sources": [ { "seller": "SourceIT (Singapore), ASUS Ascent GX10 1 TB", "trust": "Authorised reseller, local warranty, GeBIZ", "url": "https://sourceit.com.sg/products/asus-ascent-gx10-compact-desktop-ai-supercomputer-1tb-gx10-gg0007bn", "ships_from": "sg", "checked": "2026-10-09T04:33+08:00", "ok": false, "error": "HTTPError: HTTP Error 429: Too Many Requests" }, { "seller": "SourceIT (Singapore), DGX Spark 4 TB", "trust": "Authorised reseller, local warranty, GeBIZ", "url": "https://sourceit.com.sg/products/nvidia-dgx-spark-ai-supercomputer-4tb-940-54242-0007-000", "ships_from": "sg", "checked": "2026-10-09T04:33+08:00", "ok": false, "error": "HTTPError: HTTP Error 429: Too Many Requests" }, { "seller": "Vii PC Trade (Singapore), DGX Spark 4 TB", "trust": "Local reseller, 1yr warranty", "url": "https://www.viipc.sg/products/nvidia-dgx-spark-ai-server-enterprise-gpu-computing-platform-4tb-128gb-ai-desktop", "ships_from": "sg", "checked": "2026-10-09T04:33+08:00", "ok": false, "error": "HTTPError: HTTP Error 429: Too Many Requests" } ], "stale": true, "last_read": "2026-10-07T04:36+08:00" }, { "id": "mac-studio", "device": "m5-ultra", "label": "Apple Mac Studio, M5 Ultra (96 GB base)", "class": "small", "from_sgd": 7999.0, "to_sgd": 9949.0, "sources_ok": 1, "sources": [ { "seller": "Apple (SG)", "trust": "Official store, configure to order", "url": "https://www.apple.com/sg/shop/buy-mac/mac-studio", "ships_from": "sg", "checked": "2026-10-09T04:33+08:00", "ok": true, "currency": "SGD", "prices": [ 7999.0, 9949.0 ], "prices_sgd": [ 7999.0, 9949.0 ] } ], "last_read": "2026-10-09T04:33+08:00" }, { "id": "mac-studio-max", "device": "m5-max", "label": "Apple Mac Studio, M5 Max (base configs)", "class": "small", "from_sgd": 3499.0, "to_sgd": 4399.0, "sources_ok": 1, "sources": [ { "seller": "Apple (SG)", "trust": "Official store, configure to order", "url": "https://www.apple.com/sg/shop/buy-mac/mac-studio", "ships_from": "sg", "checked": "2026-10-09T04:33+08:00", "ok": true, "currency": "SGD", "prices": [ 3499.0, 4399.0 ], "prices_sgd": [ 3499.0, 4399.0 ] } ], "last_read": "2026-10-09T04:33+08:00" }, { "id": "macbook-pro-max", "device": "m5-max", "label": "Apple MacBook Pro 14 or 16 inch, M5 Max (base configs)", "class": "small", "from_sgd": 5799.0, "to_sgd": 6999.0, "sources_ok": 1, "sources": [ { "seller": "Apple (SG)", "trust": "Official store, configure to order", "url": "https://www.apple.com/sg/shop/buy-mac/macbook-pro", "ships_from": "sg", "checked": "2026-10-09T04:33+08:00", "ok": true, "currency": "SGD", "prices": [ 5799.0, 5999.0, 6699.0, 6999.0 ], "prices_sgd": [ 5799.0, 5999.0, 6699.0, 6999.0 ] } ], "last_read": "2026-10-09T04:33+08:00" }, { "id": "rtx-pro-6000", "device": "rtx-pro-6000", "label": "NVIDIA RTX PRO 6000 Blackwell 96 GB (card only)", "class": "ws", "from_sgd": 28611.08, "to_sgd": 28611.08, "sources_ok": 0, "sources": [ { "seller": "SourceIT (Singapore)", "trust": "Authorised reseller, local warranty, GeBIZ", "url": "https://sourceit.com.sg/products/nvidia-rtx-pro\u2122-6000-blackwell-workstation-edition-900-5g144-2500-000", "ships_from": "sg", "checked": "2026-10-09T04:33+08:00", "ok": false, "error": "HTTPError: HTTP Error 429: Too Many Requests" } ], "stale": true, "last_read": "2026-10-06T04:34+08:00" } ], "unavailable": [ "strix-halo-128 / ALLSTARS (Singapore): ValueError: out of stock (page still lists SGD 5,599)" ], "errors": [ "strix-halo-128 / GMKtec (EVO-X2): HTTPError: HTTP Error 429: Too Many Requests", "strix-halo-128 / Bosgame (M5): HTTPError: HTTP Error 429: Too Many Requests", "gb10-128 / SourceIT (Singapore), ASUS Ascent GX10 1 TB: HTTPError: HTTP Error 429: Too Many Requests", "gb10-128 / SourceIT (Singapore), DGX Spark 4 TB: HTTPError: HTTP Error 429: Too Many Requests", "gb10-128 / Vii PC Trade (Singapore), DGX Spark 4 TB: HTTPError: HTTP Error 429: Too Many Requests", "rtx-pro-6000 / SourceIT (Singapore): HTTPError: HTTP Error 429: Too Many Requests" ] }, "errors": [ "github Blaizzy/mlx-vlm: Remote end closed connection without response", "price strix-halo-128 / GMKtec (EVO-X2): HTTPError: HTTP Error 429: Too Many Requests", "price strix-halo-128 / Bosgame (M5): HTTPError: HTTP Error 429: Too Many Requests", "price gb10-128 / SourceIT (Singapore), ASUS Ascent GX10 1 TB: HTTPError: HTTP Error 429: Too Many Requests", "price gb10-128 / SourceIT (Singapore), DGX Spark 4 TB: HTTPError: HTTP Error 429: Too Many Requests", "price gb10-128 / Vii PC Trade (Singapore), DGX Spark 4 TB: HTTPError: HTTP Error 429: Too Many Requests", "price rtx-pro-6000 / SourceIT (Singapore): HTTPError: HTTP Error 429: Too Many Requests" ] }