Fit calculator
What runs on my hardware?
Pick your rig — every model that fits, with estimated speed and max context. Fit is exact; decode is estimated from bandwidth; prefill is shown only when measured.
24GB total
8K · capped per model
On rtx-4090 at q4_k_m, 0 of 160 models fit.
Models on rtx-4090 · 0 match · click a row for detail
| Nothing fits this rig at these settings. Try a smaller quant, more GPUs, or turn off “Fits only”. | |||||
0–0 of 0 models
Fit at q4_k_m, up to 8,192 ctx, 90% usable VRAM across 1 GPU. Decode estimated on llama.cpp. Live data.