GPUFits
GPU Compatibility Checker
Two-way GPU compatibility checker: pick a model to see which GPUs run it (single or multi-card), or pick a GPU to see which models it runs, with speed estimates.
Llama 3.2 3.21B · Q4_K_M · 4K → 3.9 GB
| GPU | Verdict | Est. speed | Ref. price |
|---|---|---|---|
| rtx-3060-12gb | ✅ Comfortable | ~137 tok/s | $180 used |
| rtx-4070-ti-super | ✅ Comfortable | ~256 tok/s | $550 used |
| rx-7900-xtx | ✅ Comfortable | ~366 tok/s | $650 used |
| rtx-3090 | ✅ Comfortable | ~357 tok/s | $1,000 used |
| rtx-4090 | ✅ Comfortable | ~385 tok/s | $1,100 used |
| mac-mini-m4-pro-48gb | ✅ Comfortable | ~104 tok/s | $1,799 MSRP |
| rtx-5090 | ✅ Comfortable | ~684 tok/s | $1,999 MSRP |
| rtx-a6000 | ✅ Comfortable | ~293 tok/s | $2,800 used |
| mac-studio-m4-max-64gb | ✅ Comfortable | ~208 tok/s | $2,899 MSRP |
| mac-studio-m3-ultra-96gb | ✅ Comfortable | ~312 tok/s | $5,299 MSRP |
| a100-80gb | ✅ Comfortable | ~778 tok/s | $15,000 MSRP |
| h100-80gb | ✅ Comfortable | ~1278 tok/s | $30,000 MSRP |
Nothing local works? Rent a cloud GPU instead
Cloud GPU dealsTheoretical estimates; real-world speed varies ±30% with framework, drivers, and CPU. Prices are manual reference prices with verification dates.