self-hosted/ai
§01·spec · /gpus

RTX 4090

nvidia40 series24GB VRAM
§02·models that run on this GPU
25 total
ModelVerticalBest speedMin VRAMWorksBenchmarks
Llama 3.1 7Bllm135tok/s24GB1check ↗
Llama 3.1 8Bllm96.12tokens/s1check ↗
LTX-Video 0.9.5video9024GB1check ↗
Llama 3.1 13Bllm78tok/s24GB1check ↗
HunyuanVideovideo75s24GB2check ↗
Llama 3.1 34Bllm42tok/s24GB1check ↗
Flux.1 Devimage18s24GB4check ↗
Llama 3.1 70Bllm18tok/s24GB1check ↗
Llama 70Bllm17.79tokens/s1check ↗
WAN 2.2video12min1check ↗
Stable Diffusion XLimage5.8it/s14GB4check ↗
CogVideoX 1.5video5min12GB1check ↗
Mochi 1video5min20GB1check ↗
LTX-2video4s16GB2check ↗
LTX-Video 2Bvideo4s12GB1check ↗
Wan 2.1video4min8GB2check ↗
Wan 2.2 14Bvideo424GB1check ↗
AnimateDiff SD 1.5video1min8GB1check ↗
LTX-2.3video1min1check ↗
Flux.2-Klein-4Bimage20GB1check ↗
Flux-Devimage1check ↗
LTX Video 2.3video18GB1check ↗
Mochi 1 Previewvideo20GB1check ↗
SD1.5image1check ↗
Wan Video 2.1 14Bvideo24GB1check ↗
§03·tested recipes
showing 6 of 20