Model benchmarks

Ministral-3-3B-Instruct-2512-GGUF-Q5_K_M local LLM performance

As of November 2026, Ministral-3-3B-Instruct-2512-GGUF-Q5_K_M runs at up to 18.2 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).

LemonadeQ5_K_M
ShareRedditX

Model size

3B

Peak speed

18.2 tok/s

Average speed

18.0 tok/s

Avg PP

1667.8 tok/s

Min memory

n/a

Max context

32,768 tokens

Avg output / run

5,644 tokens

Avg runtime / run

5m 43s

Avg quality

52.9

Benchmark runs

3

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 3 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)
Overall51.352.954.0
Agent Workflow47.963.978.4
Code Generation7.922.140.9
Role Play & Narrative63.769.072.5
Research & Analysis54.256.860.9

Performance by hardware and tool

Every hardware/tool/quantization combination Ministral-3-3B-Instruct-2512-GGUF-Q5_K_M has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
NVIDIA GeForce RTX 2050LemonadeQ5_K_M18.2 tok/s18.0 tok/sn/a32,768 tokens52.93

Benchmark runs

All 3 Ministral-3-3B-Instruct-2512-GGUF-Q5_K_M runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is Ministral-3-3B-Instruct-2512-GGUF-Q5_K_M good for coding?
In our benchmarks, Ministral-3-3B-Instruct-2512-GGUF-Q5_K_M scores 22.1/100 for coding. It runs at about 18.0 tok/s, so if you want more speed, byteshape/Qwen3.8-27B-GGUF:IQ4_XS-ASCII-Condensed is faster (~42.0 tok/s) and still scores well for coding (86.3/100).
Is Ministral-3-3B-Instruct-2512-GGUF-Q5_K_M good for agentic (tool-using) tasks?
In our benchmarks, Ministral-3-3B-Instruct-2512-GGUF-Q5_K_M scores 63.9/100 for agentic workflows. It runs at about 18.0 tok/s, so if you want more speed, granite-4.2-8b-Q4_K_L is faster (~71.6 tok/s) and still scores well for agentic workflows (92.9/100).
How fast is Ministral-3-3B-Instruct-2512-GGUF-Q5_K_M for local inference?
Across 3 community benchmark runs, Ministral-3-3B-Instruct-2512-GGUF-Q5_K_M reaches up to 18.2 tok/s and averages 18.0 tok/s, with the fastest results on NVIDIA GeForce RTX 2050.
Which tools have been used to run Ministral-3-3B-Instruct-2512-GGUF-Q5_K_M?
Benchmarks were submitted using Lemonade. Results are community-contributed and updated as new runs arrive.