Model benchmarks
ministral-3-8b-instruct-2512 local LLM performance
As of October 2026, ministral-3-8b-instruct-2512 runs at up to 6.6 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
8B
Peak speed
6.6 tok/s
Average speed
6.5 tok/s
Avg PP
618.9 tok/s
Min memory
14.4 GB
Max context
16,384 tokens
Avg output / run
7,632 tokens
Avg runtime / run
20m 44s
Avg quality
61.8
Benchmark runs
2
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 2 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 56.7 | 61.8 | 67.0 |
| Agent Workflow | 47.3 | 57.6 | 67.9 |
| Code Generation | 37.3 | 40.9 | 44.4 |
| Role Play & Narrative | 70.7 | 76.9 | 83.0 |
| Research & Analysis | 71.3 | 72.0 | 72.7 |
Performance by hardware and tool
Every hardware/tool/quantization combination ministral-3-8b-instruct-2512 has been benchmarked on, ranked by peak token generation speed. Last updated October 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon Vega Series / Radeon Vega Mobile Series | Bionic | — | 6.6 tok/s | 6.5 tok/s | 14.4 GB | 16,384 tokens | 61.8 | 2 |
Benchmark runs
All 2 ministral-3-8b-instruct-2512 runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is ministral-3-8b-instruct-2512 good for coding?
- In our benchmarks, ministral-3-8b-instruct-2512 scores 40.9/100 for coding. It runs at about 6.5 tok/s, so if you want more speed, tensorfold-qwen38-27b is faster (~25.0 tok/s) and still scores well for coding (84.3/100).
- Is ministral-3-8b-instruct-2512 good for agentic (tool-using) tasks?
- In our benchmarks, ministral-3-8b-instruct-2512 scores 57.6/100 for agentic workflows. It runs at about 6.5 tok/s, so if you want more speed, muse-glimmer:latest is faster (~33.3 tok/s) and still scores well for agentic workflows (91.6/100).
- How fast is ministral-3-8b-instruct-2512 for local inference?
- Across 2 community benchmark runs, ministral-3-8b-instruct-2512 reaches up to 6.6 tok/s and averages 6.5 tok/s, with the fastest results on AMD Radeon Vega Series / Radeon Vega Mobile Series.
- How much memory does ministral-3-8b-instruct-2512 need?
- The leanest observed configuration used about 14.4 GB of memory.
- Which tools have been used to run ministral-3-8b-instruct-2512?
- Benchmarks were submitted using Bionic. Results are community-contributed and updated as new runs arrive.