Model benchmarks
mistralai/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_M local LLM performance
As of October 2026, mistralai/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_M runs at up to 68.2 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).
Model size
8B
Peak speed
68.2 tok/s
Average speed
62.8 tok/s
Avg PP
1299.2 tok/s
Min memory
11.6 GB
Max context
65,536 tokens
Avg output / run
6,775 tokens
Avg runtime / run
1m 57s
Avg quality
71.6
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.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 69.5 | 71.6 | 73.6 |
| Agent Workflow | 72.0 | 74.0 | 76.0 |
| Code Generation | 48.8 | 56.5 | 63.6 |
| Role Play & Narrative | 84.2 | 87.1 | 89.6 |
| Research & Analysis | 63.3 | 68.8 | 75.2 |
Performance by hardware and tool
Every hardware/tool/quantization combination mistralai/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_M 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 |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 4070 | llama.cpp | Q4_K_M | 68.2 tok/s | 62.8 tok/s | 11.6 GB | 65,536 tokens | 71.6 | 3 |
Benchmark runs
All 3 mistralai/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_M runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is mistralai/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_M good for coding?
- In our benchmarks, mistralai/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_M scores 56.5/100 for coding. It runs at about 62.8 tok/s, so if you want more speed, Nex-N2.5-mini-IQ3_XXS is faster (~91.4 tok/s) and still scores well for coding (81.6/100).
- Is mistralai/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_M good for agentic (tool-using) tasks?
- In our benchmarks, mistralai/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_M scores 74.0/100 for agentic workflows. It runs at about 62.8 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).
- How fast is mistralai/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_M for local inference?
- Across 3 community benchmark runs, mistralai/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_M reaches up to 68.2 tok/s and averages 62.8 tok/s, with the fastest results on NVIDIA GeForce RTX 4070.
- How much memory does mistralai/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_M need?
- The leanest observed configuration used about 11.6 GB of memory (quantizations tested: Q4_K_M).
- Which tools have been used to run mistralai/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_M?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.