Model benchmarks
qwen3:30b-thinking local LLM performance
As of September 2026, qwen3:30b-thinking runs at up to 76.8 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
30.5B
Peak speed
76.8 tok/s
Average speed
75.3 tok/s
Avg prefill
1376.4 tok/s
Min memory
20.3 GB
Max context
65,536 tokens
Avg output / run
22,699 tokens
Avg runtime / run
5m 58s
Avg quality
70.8
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 | 70.8 | 70.8 | 70.9 |
| Agent Workflow | 74.4 | 74.9 | 75.3 |
| Code Generation | 52.3 | 58.4 | 64.5 |
| Role Play & Narrative | 76.1 | 83.6 | 91.1 |
| Research & Analysis | 65.3 | 66.4 | 67.6 |
Performance by hardware and tool
Every hardware/tool/quantization combination qwen3:30b-thinking 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 Pro | Ollama | Q4_K_M | 76.8 tok/s | 75.3 tok/s | 20.3 GB | 65,536 tokens | 70.8 | 2 |
Frequently asked questions
- Is qwen3:30b-thinking good for coding?
- In our benchmarks, qwen3:30b-thinking scores 58.4/100 for coding. It runs at about 75.3 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 qwen3:30b-thinking good for agentic (tool-using) tasks?
- In our benchmarks, qwen3:30b-thinking scores 74.9/100 for agentic workflows. It runs at about 75.3 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-Q4_K_S is faster (~165.0 tok/s) and still scores well for agentic workflows (89.3/100).
- How fast is qwen3:30b-thinking for local inference?
- Across 2 community benchmark runs, qwen3:30b-thinking reaches up to 76.8 tok/s and averages 75.3 tok/s, with the fastest results on Apple M5 Pro.
- How much memory does qwen3:30b-thinking need?
- The leanest observed configuration used about 20.3 GB of memory (quantizations tested: Q4_K_M).
- Which tools have been used to run qwen3:30b-thinking?
- Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.