Yes — Devstral Small runs on CPU only 32GB
Devstral Small (24B) needs 14GB VRAM at Q4 vs your 32GB VRAM. Estimated ~3–6 tok/s (RAM offload).
Mistral’s coding specialist. Same 24B footprint, tuned for repos and tools.
| quant | weights | + KV 8k | fits 0GB? |
|---|---|---|---|
| Q4 | 14GB | ~15GB | × tight/no |
| Q5 | 17GB | ~18GB | × tight/no |
| Q8 | 25GB | ~26GB | × tight/no |
| FP16 | 48GB | ~49GB | × tight/no |
File size ≠ runtime. KV grows with context — halve to 4k/2k if long chats OOM. Leave 1–2GB free.
Can Devstral Small run on CPU only 32GB?
Devstral Small (24B) needs 14GB VRAM at Q4 vs your 32GB VRAM. Estimated ~3–6 tok/s (RAM offload). Model context 128k. Compare with all models or re-check with your exact specs on the bench.
Unload with ollama stop, cap layers --num-gpu 28, cut context to 2k. See the OOM fixer on the homepage.
Start Q4. Bigger model at Q4 beats smaller at Q8. Only go Q8 if 2GB+ headroom remains.
About ~3–6 tok/s (RAM offload) when 100% on GPU. Any CPU offload drops speed 5–20× — check ollama ps.
Yes — Apache-2.0 allows commercial use.