self-hosted/ai
§01·spec · /gpus

RTX 3060 Ti

nvidia30 series8GB VRAM

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

§02·models that run on this GPU
22 total
LLM · 12
ModelBest speedMin VRAMWorksEvidence
llama2 7b73.07tokens/s8GB1benchrecipecheck ↗
wizardlm2 7b70.79tokens/s8GB1benchrecipecheck ↗
qwen2 7b63.73tokens/s8GB1benchrecipecheck ↗
Qwen2.5 7B58.13tokens/s8GB1benchrecipecheck ↗
Llama 3.1 8B57.34tokens/s8GB1benchrecipecheck ↗
gemma 7b31.95tokens/s8GB1benchrecipecheck ↗
falcon2 11b31.2tokens/s8GB1benchrecipecheck ↗
Gemma 2 9B23.8tokens/s8GB1benchrecipecheck ↗
stablelm2 12b18.73tokens/s8GB1benchrecipecheck ↗
llama2 13b9.25tokens/s8GB1benchrecipecheck ↗
Ornith 1.0 9B8GBrecipecheck ↗
Qwen3-4B4GBrecipecheck ↗
Multimodal · 3
ModelBest speedMin VRAMWorksEvidence
llava 7b72tokens/s8GB1benchrecipecheck ↗
Gemma 4 E4B-IT6GBrecipecheck ↗
MiniMind-O4GBrecipecheck ↗
Video · 1
ModelBest speedMin VRAMWorksEvidence
AnimateDiff6GBrecipecheck ↗
TTS · 5
specialized · 1
ModelBest speedMin VRAMWorksEvidence
SAM 34GBrecipecheck ↗
§03·tested recipes
showing 6 of 22