self-hosted/ai
§01·spec · /gpus

RTX 3060

nvidia30 series12GB VRAM

38 open-weights AI models run on the RTX 3060 — see which fit, how fast they go, and the VRAM each needs.

§02·models that run on this GPU
38 total
LLM · 6
ModelBest speedMin VRAMWorksEvidence
gpt-oss 20B64tokens/s12GB1benchrecipecheck ↗
Qwen3-8B55.2tokens/s12GB2benchesrecipecheck ↗
Llama 3.1 8B52.2tokens/s10GB2benchesrecipecheck ↗
Qwen3 14B35.9tok/s12GB3benchesrecipecheck ↗
Nanbeige4.2 3B12GBrecipecheck ↗
Ornith 1.0 9B12GBrecipecheck ↗
Multimodal · 6
ModelBest speedMin VRAMWorksEvidence
Gemma 4 E4B-IT45tokens/s6GB1benchrecipecheck ↗
Qwen3.6 35B-A3B38.9tok/s9.8GB1benchcheck ↗
Gemma 4 26B MoE37.2tok/s1benchcheck ↗
Fara1.5-4B12GBrecipecheck ↗
Fara1.5-9B12GBrecipecheck ↗
MiniMind-O4GBrecipecheck ↗
Video · 3
ModelBest speedMin VRAMWorksEvidence
LightX2V12GBrecipecheck ↗
MiniMax H3 (Hailuo 3)12GBrecipecheck ↗
WAN 2.28GBrecipecheck ↗
3D · 1
ModelBest speedMin VRAMWorksEvidence
Hunyuan3D10GBrecipecheck ↗
specialized · 2
ModelBest speedMin VRAMWorksEvidence
KiMoDo3GBrecipecheck ↗
SAM 34GBrecipecheck ↗
§03·tested recipes
showing 6 of 36