Written by Jakub Rusinowski · Last updated August 15, 2026
Yes
Yes — Devstral Small 24B at Q8_0 needs about 27.6 GB of the 32 GB usable on RTX 5090 Desktop (32 GB VRAM, 64 GB RAM) (~4.4 GB spare), at ~46 tok/s (estimated), with room for about 32,768 tokens of context.
Confidence: high · Recommended quantization: Q8_0 · Estimated speed: ~46 tok/s
| Usable memory for models | 32 GB |
| Memory bandwidth | 1792 GB/s |
| Form factor | Desktop |
| Operating system | Windows or Linux |
| Memory upgradeable | Yes |
| Quant | Memory needed | Fits 32 GB? | Max context | Est. speed | Download |
|---|---|---|---|---|---|
| F16 | 50.1 GB | ✗ No | — | — | 48 GB |
| Q8_0 | 27.6 GB | ✓ Yes | 32K | ~46 tok/s | 25.5 GB |
| Q6_K | 21.8 GB | ✓ Yes | 64K | ~57.5 tok/s | 19.7 GB |
| Q5_K_M | 19.2 GB | ✓ Yes | 64K | ~64.9 tok/s | 17 GB |
| Q4_K_M | 16.6 GB | ✓ Yes | 64K | ~73.9 tok/s | 14.5 GB |
| Q3_K_M | 12.4 GB | ✓ Yes | 64K | ~96.6 tok/s | 10.2 GB |
| Q2_K | 10 GB | ✓ Yes | 128K | ~116.1 tok/s | 7.9 GB |
| Variant | Needs at Q4_K_M | Fits? | Est. speed |
|---|---|---|---|
| Devstral-2 123B | 77.9 GB | ✗ Too large | — |
| Devstral Small 24B | 16.6 GB | ✓ Fits | ~73.9 tok/s |
| Devstral-2 22B | 15.7 GB | ✓ Fits | ~78.6 tok/s |
Ollama or llama.cpp (CUDA/ROCm) — vLLM if you need concurrent serving
Yes — Devstral Small 24B at Q8_0 needs about 27.6 GB of the 32 GB usable on RTX 5090 Desktop (32 GB VRAM, 64 GB RAM) (~4.4 GB spare), at ~46 tok/s (estimated), with room for about 32,768 tokens of context.
Q8_0 — it needs about 27.6 GB of the 32 GB available, downloads as roughly 25.5 GB, and runs at an estimated 46 tokens/sec with up to 32K 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