How much VRAM does it need?
Every time a new open model drops, the same question comes up: will it run on my GPU? We answer it for one model per page. You get the memory needed at each quantization, the GPUs and Macs it fits on, and what renting a GPU for it costs, with every number linked to its source.
How much VRAM do I need?
Pick a model or workload and your graphics card. The answer updates as you type.
Estimates. LLM numbers use real download sizes from Hugging Face plus the KV cache, the same as our VRAM table and model pages. Checked 2026-10-23.
Model pages
How every page works
- Quick answer first. One paragraph with the number you came for.
- Real file sizes. Taken from the actual downloads on Hugging Face, not guessed from parameter counts.
- Context counts. The KV cache is calculated from the model's own config, because a long context can double the memory you need.
- Sources and a "checked on" date on every page. If we can't source a number, we leave it out.