Model benchmarks
Qwen3.8-27B-MTPLX-Optimized-Quality local LLM performance
As of September 2026, Qwen3.8-27B-MTPLX-Optimized-Quality runs at up to 44.8 tok/s for local inference (best of 4 community benchmark runs across 1 GPU).
Model size
27B
Peak speed
44.8 tok/s
Average speed
39.4 tok/s
Avg PP
492.4 tok/s
Min memory
32.7 GB
Max context
262,144 tokens
Avg output / run
50,259 tokens
Avg runtime / run
23m 44s
Avg quality
85.9
Benchmark runs
4
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 4 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 | 82.3 | 85.9 | 89.2 |
| Agent Workflow | 80.7 | 86.2 | 91.0 |
| Code Generation | 73.5 | 79.4 | 84.3 |
| Role Play & Narrative | 87.9 | 91.6 | 95.9 |
| Research & Analysis | 81.6 | 86.5 | 89.3 |
Performance by hardware and tool
Every hardware/tool/quantization combination Qwen3.8-27B-MTPLX-Optimized-Quality has been benchmarked on, ranked by peak token generation speed. Last updated September 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| Apple M5 Max | oMLX | — | 44.8 tok/s | 39.4 tok/s | 32.7 GB | 262,144 tokens | 85.9 | 4 |
Benchmark runs
All 4 Qwen3.8-27B-MTPLX-Optimized-Quality runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Qwen3.8-27B-MTPLX-Optimized-Quality good for coding?
- In our benchmarks, Qwen3.8-27B-MTPLX-Optimized-Quality scores 79.4/100 for coding. It runs at about 39.4 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-Quality good for agentic (tool-using) tasks?
- In our benchmarks, Qwen3.8-27B-MTPLX-Optimized-Quality scores 86.2/100 for agentic workflows. It runs at about 39.4 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-Quality for local inference?
- Across 4 community benchmark runs, Qwen3.8-27B-MTPLX-Optimized-Quality reaches up to 44.8 tok/s and averages 39.4 tok/s, with the fastest results on Apple M5 Max.
- How much memory does Qwen3.8-27B-MTPLX-Optimized-Quality need?
- The leanest observed configuration used about 32.7 GB of memory.
- Which tools have been used to run Qwen3.8-27B-MTPLX-Optimized-Quality?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.