Model benchmarks
Qwen3-Coder-30B-A3B-Instruct-MLX-8bit local LLM performance
As of August 2026, Qwen3-Coder-30B-A3B-Instruct-MLX-8bit runs at up to 90.0 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
30B
Peak speed
90.0 tok/s
Average speed
89.8 tok/s
Min memory
32.1 GB
Max context
65,536 tokens
Avg runtime / run
1m 38s
Avg quality
48.7
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 | 43.9 | 48.7 | 53.4 |
| Agent Workflow | 16.2 | 24.1 | 32.0 |
| Code Generation | 55.9 | 59.0 | 62.2 |
| Role Play & Narrative | 54.3 | 65.1 | 76.0 |
| Research & Analysis | 43.1 | 46.5 | 49.9 |
Performance by hardware and tool
Every hardware/tool/quantization combination Qwen3-Coder-30B-A3B-Instruct-MLX-8bit 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 | — | 90.0 tok/s | 89.8 tok/s | 32.1 GB | 65,536 tokens | 48.7 | 2 |
Benchmark runs
All 2 Qwen3-Coder-30B-A3B-Instruct-MLX-8bit runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Qwen3-Coder-30B-A3B-Instruct-MLX-8bit good for coding?
- In our benchmarks, Qwen3-Coder-30B-A3B-Instruct-MLX-8bit scores 59.0/100 for coding. It runs at about 89.8 tok/s, so if you want more speed, unsloth/Qwen3.8-27B-GGUF:IQ3_S is faster (~104.3 tok/s) and still scores well for coding (83.0/100).
- Is Qwen3-Coder-30B-A3B-Instruct-MLX-8bit good for agentic (tool-using) tasks?
- In our benchmarks, Qwen3-Coder-30B-A3B-Instruct-MLX-8bit scores 24.1/100 for agentic workflows. It runs at about 89.8 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-Q4_K_S is faster (~162.4 tok/s) and still scores well for agentic workflows (87.8/100).
- How fast is Qwen3-Coder-30B-A3B-Instruct-MLX-8bit for local inference?
- Across 2 community benchmark runs, Qwen3-Coder-30B-A3B-Instruct-MLX-8bit reaches up to 90.0 tok/s and averages 89.8 tok/s, with the fastest results on Apple M5 Max.
- How much memory does Qwen3-Coder-30B-A3B-Instruct-MLX-8bit need?
- The leanest observed configuration used about 32.1 GB of memory.
- Which tools have been used to run Qwen3-Coder-30B-A3B-Instruct-MLX-8bit?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.