About GPUFits

GPUFits answers one question: can your GPU run this LLM, and how well? We maintain verified datasets of popular open-weight models, consumer GPUs, quantization formats, and API prices — then turn them into calculators and straight answers.

Our data methodology

Every number on this site carries a source and a verification date:

• Model specs (parameters, layers, KV heads, context length) come from official Hugging Face configs.
• GPU specs (VRAM, bandwidth) come from NVIDIA, AMD, and Apple official specifications. Used prices are sampled from eBay sold listings and marked as dated snapshots.
• Quantization sizes are calibrated from measured GGUF file sizes, not theoretical bit-rates.
• API prices are taken from official provider pricing pages.

All datasets are re-verified on the first weekend of every month. Every change is logged on our data updates page.

What our estimates are (and are not)

VRAM requirements are computed from published architecture constants and measured file sizes — they match real-world usage closely, but runtimes differ. Speed figures are theoretical estimates with a stated ±30% error band. Nothing here is purchasing advice; see our disclaimer.

Contact

Found a stale number? Data corrections are our favorite email — see the contact page.