Compatibility Matrix

Best community run per VRAM tier for the most-run open-weight models. Cells show tok/s and quantization — the same model often lands in multiple tiers via different quants. Full board · Submit your run.

Model familyAll Arcee Baichuan BTL Cohere Deepseek Deepseek-Coder Deepseek-R1 DeepSeek-V4.1-Flash Gemma GLM Gpt Hunyuan Kimi Kwaipilot Laguna LFM Ling Llama MiniCPM-V Minimax Mistral MobileLLM Muse Nemotron Ornith Phi Qwen Qwen3.8 RNJ Spark unknown

Model	Family	<24GB	24–48GB	>48GB	runs
DeepSeek-V4.1-Flash	Deepseek	—	26.0 · EXL3 3.0 bpw	231.6 · FP4+FP8 native	39
DeepSeek-V4-Flash-0731	Deepseek	—	91.2 · FP8_E4M3	185.6 · FP8	7
DeepSeek-V4.1-Flash-1017	DeepSeek-V4.1-Flash	—	—	190.0 · EXL3 3.5bpw	3
DeepSeek-R1-Distill-Qwen-7B	Deepseek-R1	39.8 · Q8_0	213.3 · Q4_K_M	—	2
DeepSeek-V4-Pro-Qwen3.5-9B-MTP	Deepseek	—	43.4 · Q8_0	—	1
DeepSeek-V4-Flash	Deepseek	—	53.1 · MQ2R	—	1
DeepSeek-Coder-V2-Lite-Instruct	Deepseek-Coder	—	—	237.3 · FP8	1
DeepSeek-V4-Flash-180B	Deepseek	—	80.8 · FP8	—	1
DeepSeek-R1-0528-Qwen3-8B	Deepseek-R1	—	141.5 · Q8_0	—	1
DeepSeek-V4-Flash-0731-ROCmFP3	Deepseek	—	84.8 · Q4_K_M	—	1
DeepSeek-R1-Distill-Qwen-14B	Deepseek-R1	35.1 · Q4_K_M	—	—	1
keys-DeepSeekV4Flash-Vision-EXP-ablit	Deepseek	—	66.8 · FP8	—	1
DeepSeek-V4-Flash-Vision-Exp	Deepseek	—	—	179.5 · FP8	1

60 runs · last run Oct 17, 2026. Filter by family above; full board for every row.

VRAM tiers: <24 GB · 24–48 GB · >48 GB. A cell shows the best single-stream tok/s and quant recorded in that tier — not a guarantee for every loadout. Imported rows carry their original methodology link.

Related guides: best GPU for Qwen3 · quantization · Strix Halo 128GB