Model benchmarks

mistral-small3.2:latest local LLM performance

As of April 2026, mistral-small3.2:latest runs at up to 32.5 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).

OllamaQ4_K_M

Model size

24.0B

Peak speed

32.5 tok/s

Average speed

22.0 tok/s

Min memory

20.0 GB

Max context

5,068 tokens

Best quality

85.1

Benchmark runs

2

GPUs tested

1

Performance by hardware and tool

Every hardware/tool/quantization combination mistral-small3.2: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_M32.5 tok/s22.0 tok/s20.0 GB5,068 tokens69.42

Frequently asked questions

How fast is mistral-small3.2:latest for local inference?
Across 2 community benchmark runs, mistral-small3.2:latest reaches up to 32.5 tok/s and averages 22.0 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
How much memory does mistral-small3.2:latest need?
The leanest observed configuration used about 20.0 GB of memory (quantizations tested: Q4_K_M).
Which tools have been used to run mistral-small3.2:latest?
Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.