Model benchmarks

devstral-small-2:latest local LLM performance

As of April 2026, devstral-small-2:latest runs at up to 28.5 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).

OllamaQ4_K_M

Model size

24.0B

Peak speed

28.5 tok/s

Average speed

16.9 tok/s

Min memory

16.3 GB

Max context

4,334 tokens

Best quality

93.1

Benchmark runs

3

GPUs tested

1

Performance by hardware and tool

Every hardware/tool/quantization combination devstral-small-2:latest has been benchmarked on, ranked by peak token generation speed. Last updated April 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
AMD Radeon RX 7900 XTXOllamaQ4_K_M28.5 tok/s16.9 tok/s16.3 GB4,334 tokens63.23

Frequently asked questions

How fast is devstral-small-2:latest for local inference?
Across 3 community benchmark runs, devstral-small-2:latest reaches up to 28.5 tok/s and averages 16.9 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
How much memory does devstral-small-2:latest need?
The leanest observed configuration used about 16.3 GB of memory (quantizations tested: Q4_K_M).
Which tools have been used to run devstral-small-2:latest?
Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.