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
NVIDIA-Nemotron-3.5-Lightning-30B-A3B	Nemotron	116.0 · NVFP4	446.7 · NVFP4	217.9 · W4A16	9
NVIDIA-Nemotron-3-Nano-4B	Nemotron	185.8 · Q4_K_M	237.8 · Q8_0	—	3
Nemotron-3-Nano-Omni-30B-A3B-Reasoning	Nemotron	—	274.4 · Q4_K_M	—	1
NVIDIA-Nemotron-3-Nano-30B-A3B	Nemotron	—	286.8 · Q4_K_M	—	1
OpenReasoning-Nemotron-7B	Nemotron	68.0 · Q4_K_M	—	—	1
NVIDIA-Nemotron-Nano-9B-v2	Nemotron	64.2 · IQ4_XS	—	—	1
NVIDIA-Nemotron-Nano-12B-v2	Nemotron	40.4 · Q4_K_M	—	—	1
Nemotron-Cascade-2-30B-A3B	Nemotron	—	95.4 · Q8_0	—	1
NVIDIA-Nemotron-3.5-Super-VL-120B-A12B	Nemotron	—	—	48.0 · NVFP4	1

19 runs · last run Oct 16, 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