self-hosted/ai
§01·spec · /gpus

RTX 5090

nvidia50 series32GB VRAM

40 open-weights AI models run on the RTX 5090 — see which fit, how fast they go, and the VRAM each needs.

§02·models that run on this GPU
40 total
LLM · 14
Multimodal · 5
ModelBest speedMin VRAMWorksEvidence
Gemma 4 26B MoE180.3tokens/s29GB2benchesrecipecheck ↗
Qwen3.5 35B165.2tokens/s24GB2benchesrecipecheck ↗
Gemma4 31B61.1tokens/s20GB2benchesrecipecheck ↗
Qwen3.5 27B58.8tokens/s20GB2benchesrecipecheck ↗
Gemma 4 E4B-IT20GBrecipecheck ↗
3D · 4
ModelBest speedMin VRAMWorksEvidence
Hunyuan3D29GBrecipecheck ↗
TRELLIS.2-4B24GBrecipecheck ↗
TRELLIS image-large16GBrecipecheck ↗
Waypoint 1.512GBrecipecheck ↗
TTS · 1
ModelBest speedMin VRAMWorksEvidence
Kokoro TTS3GBrecipecheck ↗
§03·tested recipes
showing 6 of 40