Model benchmarks
Qwen3.6-27B-MLX-4bit local LLM performance
As of August 2026, Qwen3.6-27B-MLX-4bit runs at up to 32.1 tok/s for local inference (best of 5 community benchmark runs across 1 GPU).
oMLX
Model size
27B
Peak speed
32.1 tok/s
Average speed
20.2 tok/s
Min memory
n/a
Max context
65,536 tokens
Avg runtime / run
8m 58s
Avg quality
82.9
Benchmark runs
5
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 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 | 80.8 | 82.9 | 85.3 |
| Agent Workflow | 76.4 | 85.6 | 90.6 |
| Code Generation | 74.2 | 77.3 | 79.2 |
| Role Play & Narrative | 81.2 | 86.0 | 92.3 |
| Research & Analysis | 80.1 | 82.8 | 85.9 |
Performance by hardware and tool
Every hardware/tool/quantization combination Qwen3.6-27B-MLX-4bit 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 | — | 32.1 tok/s | 20.2 tok/s | n/a | 65,536 tokens | 82.9 | 5 |
Frequently asked questions
- Is Qwen3.6-27B-MLX-4bit good for coding?
- In our benchmarks, Qwen3.6-27B-MLX-4bit scores 77.3/100 for coding. It runs at about 20.2 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-MLX-oQ4e-MTP is faster (~123.6 tok/s) and still scores well for coding (76.5/100).
- Is Qwen3.6-27B-MLX-4bit good for agentic (tool-using) tasks?
- In our benchmarks, Qwen3.6-27B-MLX-4bit scores 85.6/100 for agentic workflows. It runs at about 20.2 tok/s, so if you want more speed, openai/gpt-oss-20b:2 is faster (~154.8 tok/s) and still scores well for agentic workflows (87.2/100).
- How fast is Qwen3.6-27B-MLX-4bit for local inference?
- Across 5 community benchmark runs, Qwen3.6-27B-MLX-4bit reaches up to 32.1 tok/s and averages 20.2 tok/s, with the fastest results on Apple M5 Max.
- How much memory does Qwen3.6-27B-MLX-4bit need?
- The leanest observed configuration used about n/a of memory.
- Which tools have been used to run Qwen3.6-27B-MLX-4bit?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.