AI Mo Kimina Prover 72B speed on RTX 5090 and quantization-level VRAM fit.
RTX 5090 does not meet the minimum VRAM requirement for Q4 inference of AI Mo Kimina Prover 72B. Review the quantization breakdown below to see how higher precision settings impact VRAM and throughput.
RTX 5090 lacks sufficient VRAM for comfortable AI Mo Kimina Prover 72B operation with Q4 quantization.
Your 32GB GPU is 4GB short of the 36GB minimum.
Options: (1) Try Q2 or Q3 quantization for lower VRAM requirements, (2) Consider cloud GPU rental, (3) Upgrade to a GPU with at least 16GB VRAM.
| Quantization | VRAM needed | VRAM available | Estimated speed | Verdict |
|---|---|---|---|---|
| Q4 | 36GB | 32GB | 59.95 tok/s | ❌ Not recommended |
| Q8 | 72GB | 32GB | 41.97 tok/s | ❌ Not recommended |
| FP16 | 144GB | 32GB | 22.78 tok/s | ❌ Not recommended |
Check current pricing links for RTX 5090 and similar cards.
Open RTX 5090 buy links →Use workload-focused recommendations before committing to a purchase.
Browse best GPU guides →Compare complete systems if you want ready-to-run hardware.
Compare prebuilt systems →RTX 5090 is not a comfortable Q4 fit for AI Mo Kimina Prover 72B (about 36GB needed).
Q4 inference is estimated to need about 36GB VRAM on this page, while RTX 5090 has 32GB available.
Try lower-bit quantization, choose a smaller model, or move to a higher-VRAM GPU from the alternatives list.