Model benchmarks
Qwen3.8-27B-MTPLX-Optimized-Speed local LLM performance
As of October 2026, Qwen3.8-27B-MTPLX-Optimized-Speed runs at up to 52.7 tok/s for local inference (best of 5 community benchmark runs across 2 GPUs).
Model size
27B
Peak speed
52.7 tok/s
Average speed
35.0 tok/s
Avg PP
308.6 tok/s
Min memory
24.4 GB
Max context
131,072 tokens
Avg output / run
16,180 tokens
Avg runtime / run
7m 59s
Avg quality
79.9
Benchmark runs
5
GPUs tested
2
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 5 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 | 68.7 | 79.9 | 87.7 |
| Agent Workflow | 80.0 | 83.7 | 87.8 |
| Code Generation | 44.4 | 68.9 | 84.0 |
| Role Play & Narrative | 75.3 | 84.4 | 94.0 |
| Research & Analysis | 74.5 | 82.5 | 88.5 |
Performance by hardware and tool
Every hardware/tool/quantization combination Qwen3.8-27B-MTPLX-Optimized-Speed 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 |
|---|---|---|---|---|---|---|---|---|
| Apple M5 Max | oMLX | — | 52.7 tok/s | 51.5 tok/s | 26.2 GB | 131,072 tokens | 84.2 | 2 |
| Apple M4 Pro | oMLX | — | 25.0 tok/s | 24.1 tok/s | 24.4 GB | 32,768 tokens | 77.0 | 3 |
Benchmark runs
All 5 Qwen3.8-27B-MTPLX-Optimized-Speed runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Qwen3.8-27B-MTPLX-Optimized-Speed good for coding?
- In our benchmarks, Qwen3.8-27B-MTPLX-Optimized-Speed scores 68.9/100 for coding. It runs at about 35.0 tok/s, so if you want more speed, local-qwen is faster (~60.2 tok/s) and still scores well for coding (84.8/100).
- Is Qwen3.8-27B-MTPLX-Optimized-Speed good for agentic (tool-using) tasks?
- In our benchmarks, Qwen3.8-27B-MTPLX-Optimized-Speed scores 83.7/100 for agentic workflows. It runs at about 35.0 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 Qwen3.8-27B-MTPLX-Optimized-Speed for local inference?
- Across 5 community benchmark runs, Qwen3.8-27B-MTPLX-Optimized-Speed reaches up to 52.7 tok/s and averages 35.0 tok/s, with the fastest results on Apple M5 Max.
- How much memory does Qwen3.8-27B-MTPLX-Optimized-Speed need?
- The leanest observed configuration used about 24.4 GB of memory.
- Which tools have been used to run Qwen3.8-27B-MTPLX-Optimized-Speed?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.