Model benchmarks

ministral-3:14b local LLM performance

As of April 2026, ministral-3:14b runs at up to 47.0 tok/s for local inference (best of 12 community benchmark runs across 1 GPU).

OllamaQ4_K_M

Model size

13.9B

Peak speed

47.0 tok/s

Average speed

43.3 tok/s

Min memory

10.6 GB

Max context

5,481 tokens

Best quality

91.4

Benchmark runs

12

GPUs tested

1

Performance by hardware and tool

Every hardware/tool/quantization combination ministral-3:14b 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_M47.0 tok/s43.3 tok/s10.6 GB5,481 tokens67.412

Frequently asked questions

How fast is ministral-3:14b for local inference?
Across 12 community benchmark runs, ministral-3:14b reaches up to 47.0 tok/s and averages 43.3 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
How much memory does ministral-3:14b need?
The leanest observed configuration used about 10.6 GB of memory (quantizations tested: Q4_K_M).
Which tools have been used to run ministral-3:14b?
Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.