Model benchmarks

ministral-3:14b local LLM performance

As of April 2026, ministral-3:14b runs at up to 43.0 tok/s for local inference (best of 6 community benchmark runs across 1 GPU).

OllamaQ4_K_M
ShareRedditX

Model size

13.9B

Peak speed

43.0 tok/s

Average speed

42.3 tok/s

Min memory

14.4 GB

Max context

65,536 tokens

Avg quality

66.8

Benchmark runs

6

GPUs tested

1

Quality by task

Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 6 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.

TaskP5 (low)AvgP95 (high)
Overall59.766.872.1
Agent Workflow63.972.379.8
Code Generation48.854.660.2
Role Play & Narrative67.674.281.6
Research & Analysis53.965.976.1

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_M43.0 tok/s42.3 tok/s14.4 GB65,536 tokens66.76

Frequently asked questions

Is ministral-3:14b good for coding?
In our benchmarks, ministral-3:14b scores 54.6/100 for coding. It runs at about 42.3 tok/s, so if you want more speed, Gemma4:E2B/QAT-MTP@131K is faster (~303.9 tok/s) and still scores well for coding (66.9/100).
Is ministral-3:14b good for agentic (tool-using) tasks?
In our benchmarks, ministral-3:14b scores 72.3/100 for agentic workflows. It runs at about 42.3 tok/s, so if you want more speed, Gemma4:E2B/QAT-MTP@131K is faster (~303.9 tok/s) and still scores well for agentic workflows (71.7/100).
How fast is ministral-3:14b for local inference?
Across 6 community benchmark runs, ministral-3:14b reaches up to 43.0 tok/s and averages 42.3 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 14.4 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.