Written by Jakub Rusinowski · Last updated January 28, 2025
Yes, but it is tight
Yes, but it is tight — Mistral Small 3 (24B) at Q6_K needs about 22.4 GB of the 24 GB usable on RTX 4090 Desktop (24 GB VRAM, 64 GB RAM), leaving only ~1.6 GB before the runtime starts swapping. Expect ~33.9 tok/s (estimated), with room for about 8,192 tokens of context.
Confidence: medium · Recommended quantization: Q6_K · Estimated speed: ~33.9 tok/s
| Usable memory for models | 24 GB |
| Memory bandwidth | 1008 GB/s |
| Form factor | Desktop |
| Operating system | Windows or Linux |
| Memory upgradeable | Yes |
| Quant | Memory needed | Fits 24 GB? | Max context | Est. speed | Download |
|---|---|---|---|---|---|
| F16 | 50.7 GB | ✗ No | — | — | 48 GB |
| Q8_0 | 28.2 GB | ✗ No | — | — | 25.5 GB |
| Q6_K | 22.4 GB | ✓ Yes | 8K | ~33.9 tok/s | 19.7 GB |
| Q5_K_M | 19.7 GB | ✓ Yes | 16K | ~38.5 tok/s | 17 GB |
| Q4_K_M | 17.2 GB | ✓ Yes | 32K | ~44.1 tok/s | 14.5 GB |
| Q3_K_M | 12.9 GB | ✓ Yes | 32K | ~58.8 tok/s | 10.2 GB |
| Q2_K | 10.6 GB | ✓ Yes | 32K | ~71.8 tok/s | 7.9 GB |
| Variant | Needs at Q4_K_M | Fits? | Est. speed |
|---|---|---|---|
| Mistral Small 3 (24B) | 17.2 GB | ✓ Fits | ~44.1 tok/s |
| Mistral NeMo 12B | 9.4 GB | ✓ Fits | ~78.6 tok/s |
Ollama or llama.cpp (CUDA/ROCm) — vLLM if you need concurrent serving
Yes, but it is tight — Mistral Small 3 (24B) at Q6_K needs about 22.4 GB of the 24 GB usable on RTX 4090 Desktop (24 GB VRAM, 64 GB RAM), leaving only ~1.6 GB before the runtime starts swapping. Expect ~33.9 tok/s (estimated), with room for about 8,192 tokens of context.
Q6_K — it needs about 22.4 GB of the 24 GB available, downloads as roughly 19.7 GB, and runs at an estimated 33.9 tokens/sec with up to 8K of context.
Nothing binding — the model fits with headroom and generates at a usable speed on this hardware.
Ollama or llama.cpp (CUDA/ROCm) — vLLM if you need concurrent serving
← Can I Run It? | Mistral Family model page | Check your hardware