Model benchmarks
Qwen3.8-27B-Uncensored-MLX local LLM performance
As of August 2026, Qwen3.8-27B-Uncensored-MLX runs at up to 32.6 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
27B
Peak speed
32.6 tok/s
Average speed
31.1 tok/s
Min memory
17.8 GB
Max context
262,144 tokens
Avg runtime / run
15m 44s
Avg quality
76.8
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 | 68.7 | 76.8 | 84.9 |
| Agent Workflow | 88.3 | 88.9 | 89.5 |
| Code Generation | 16.5 | 44.3 | 72.0 |
| Role Play & Narrative | 90.0 | 92.2 | 94.4 |
| Research & Analysis | 78.7 | 81.9 | 85.0 |
Performance by hardware and tool
Every hardware/tool/quantization combination Qwen3.8-27B-Uncensored-MLX has been benchmarked on, ranked by peak token generation speed. Last updated August 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| Apple M5 Max | oMLX | — | 32.6 tok/s | 31.1 tok/s | 17.8 GB | 262,144 tokens | 76.8 | 2 |
Benchmark runs
All 2 Qwen3.8-27B-Uncensored-MLX runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Qwen3.8-27B-Uncensored-MLX good for coding?
- In our benchmarks, Qwen3.8-27B-Uncensored-MLX scores 44.3/100 for coding. It runs at about 31.1 tok/s, so if you want more speed, Qwen3.8-27B-oQ4e-fp16-mtp is faster (~38.6 tok/s) and still scores well for coding (82.6/100).
- Is Qwen3.8-27B-Uncensored-MLX good for agentic (tool-using) tasks?
- In our benchmarks, Qwen3.8-27B-Uncensored-MLX scores 88.9/100 for agentic workflows. It runs at about 31.1 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 Qwen3.8-27B-Uncensored-MLX for local inference?
- Across 2 community benchmark runs, Qwen3.8-27B-Uncensored-MLX reaches up to 32.6 tok/s and averages 31.1 tok/s, with the fastest results on Apple M5 Max.
- How much memory does Qwen3.8-27B-Uncensored-MLX need?
- The leanest observed configuration used about 17.8 GB of memory.
- Which tools have been used to run Qwen3.8-27B-Uncensored-MLX?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.