Model benchmarks
Qwen3.8-27B-oQ6e-mtp local LLM performance
As of September 2026, Qwen3.8-27B-oQ6e-mtp runs at up to 49.1 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
27B
Peak speed
49.1 tok/s
Average speed
39.1 tok/s
Avg PP
500.2 tok/s
Min memory
26.3 GB
Max context
262.144 tokens
Avg output / run
37.042 tokens
Avg runtime / run
18m 32s
Avg quality
86.1
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 | 85.8 | 86.1 | 86.5 |
| Agent Workflow | 86.5 | 88.6 | 90.7 |
| Code Generation | 76.5 | 77.3 | 78.1 |
| Role Play & Narrative | 91.1 | 93.4 | 95.6 |
| Research & Analysis | 84.8 | 85.3 | 85.8 |
Performance by hardware and tool
Every hardware/tool/quantization combination Qwen3.8-27B-oQ6e-mtp 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 | oQ6e | 49.1 tok/s | 49.1 tok/s | 30.4 GB | 262.144 tokens | 86.6 | 1 |
| Apple M5 Max | oMLX | — | 29.2 tok/s | 29.2 tok/s | 26.3 GB | 262.144 tokens | 85.7 | 1 |
Benchmark runs
All 2 Qwen3.8-27B-oQ6e-mtp runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Qwen3.8-27B-oQ6e-mtp good for coding?
- In our benchmarks, Qwen3.8-27B-oQ6e-mtp scores 77.3/100 for coding. It runs at about 39.1 tok/s, so if you want more speed, ornith-ai/Ornith-1.5-35B-A3B-GGUF is faster (~96.3 tok/s) and still scores well for coding (83.4/100).
- Is Qwen3.8-27B-oQ6e-mtp good for agentic (tool-using) tasks?
- In our benchmarks, Qwen3.8-27B-oQ6e-mtp scores 88.6/100 for agentic workflows. It runs at about 39.1 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-oQ6e-mtp for local inference?
- Across 2 community benchmark runs, Qwen3.8-27B-oQ6e-mtp reaches up to 49.1 tok/s and averages 39.1 tok/s, with the fastest results on Apple M5 Max.
- How much memory does Qwen3.8-27B-oQ6e-mtp need?
- The leanest observed configuration used about 26.3 GB of memory (quantizations tested: oQ6e).
- Which tools have been used to run Qwen3.8-27B-oQ6e-mtp?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.