Model benchmarks
qwen/qwen3-coder-30b local LLM performance
As of August 2026, qwen/qwen3-coder-30b runs at up to 240.2 tok/s for local inference (best of 2 community benchmark runs across 2 GPUs).
Model size
30B
Peak speed
240.2 tok/s
Average speed
162.4 tok/s
Min memory
16.0 GB
Max context
32,768 tokens
Avg output / run
7,437 tokens
Avg runtime / run
1m 36s
Avg quality
55.4
Benchmark runs
2
GPUs tested
2
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 | 54.5 | 55.4 | 56.3 |
| Agent Workflow | 41.6 | 43.8 | 46.1 |
| Code Generation | 56.9 | 58.7 | 60.4 |
| Role Play & Narrative | 61.2 | 61.7 | 62.1 |
| Research & Analysis | 57.6 | 57.6 | 57.7 |
Performance by hardware and tool
Every hardware/tool/quantization combination qwen/qwen3-coder-30b 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 |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 5090 | Ollama | Q4_K_M | 240.2 tok/s | 240.2 tok/s | 18.9 GB | 32,768 tokens | 56.4 | 1 |
| Apple M5 Pro | LM Studio | — | 84.5 tok/s | 84.5 tok/s | 16.0 GB | 32,768 tokens | 54.5 | 1 |
Frequently asked questions
- Is qwen3-coder:30b good for coding?
- In our benchmarks, qwen3-coder:30b scores 58.7/100 for coding. It runs at about 162.4 tok/s, so if you want more speed, Gemma4:E2B/QAT-MTP@131K is faster (~303.9 tok/s) and still scores well for coding (66.9/100).
- Is qwen3-coder:30b good for agentic (tool-using) tasks?
- In our benchmarks, qwen3-coder:30b scores 43.8/100 for agentic workflows. It runs at about 162.4 tok/s, so if you want more speed, Gemma4:E2B/QAT-MTP@131K is faster (~303.9 tok/s) and still scores well for agentic workflows (71.7/100).
- How fast is qwen/qwen3-coder-30b for local inference?
- Across 2 community benchmark runs, qwen/qwen3-coder-30b reaches up to 240.2 tok/s and averages 162.4 tok/s, with the fastest results on NVIDIA GeForce RTX 5090.
- How much memory does qwen/qwen3-coder-30b need?
- The leanest observed configuration used about 16.0 GB of memory (quantizations tested: Q4_K_M).
- Which tools have been used to run qwen/qwen3-coder-30b?
- Benchmarks were submitted using LM Studio, Ollama. Results are community-contributed and updated as new runs arrive.