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
gemma-4-26B-A4B-it	Gemma	62.1 · UD-Q4_K_M	209.8 · Q4_K_M	—	4
gemma-4-31B-it	Gemma	15.8 · Q4_K_M	25.4 · Q4_K_XL	28.3 · NVFP4	4
gemma-4-12B-it	Gemma	89.7 · Q4_K_XL	88.3 · Q8_0	—	4
gemma-4-12B-it-qat	Gemma	71.2 · Q4_0	130.8 · Q4_0	—	3
gemma-4-E4B-it	Gemma	70.9 · Q4_K_M	161.5 · Q8_0	—	3
gemma-4-E2B-it	Gemma	113.4 · Q4_K_M	232.6 · Q4_K_M	—	3
gemma-4-26B-A4B-it-qat	Gemma	95.3 · Q4_K_M	254.3 · Q4_0	—	2
gemma-4-26B-A4B-it-ultra-uncensored-heretic	Gemma	88.2 · Q4_K_S	—	—	1
gemma-3-1b-it	Gemma	—	—	664.3 · Q4_K_M	1
Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-MTP	Gemma	—	58.4 · Q4_K_M	—	1
gemma-3n-E4B-it-text	Gemma	—	135.4 · Q8_0	—	1
gemma-3-12b-it	Gemma	39.4 · Q4_K_M	—	—	1

28 runs · last run Sep 4, 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