AlibabaQwen3 30B A3B
Best configuration ranks #389 of 610 on the BenchLeader Index at 45.4 ±5.0. Last measured 2 Sept 2026. Released 29 Apr 2025.
- Blended price
- $0.215/M
- $0.120 in · $0.500 out
- Output speed
- 108 tok/s
- First answer
- 2.26 s
- first token 0.68 s
- Context
- 131k
- Overall index45
- Reasoning50
- Coding47
- Agents & tools48
- Maths52
- Knowledge33
- Instruction following36
- Human preference56
- Long context38
- Composite38
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoning | – | – | 143 tok/s | 1.98 s | $0.0002 | Maths 35 · Reasoning 38 |
| thinking | 45.1 | #397 | 148 tok/s | 21 s | $0.0008 | Agents & tools 38 · Coding 35 · Composite 41 · Instruction following 52 · Knowledge 40 · Long context 57 · Maths 38 · Reasoning 51 |
| defaultbest | 45.4 | #389 | 108 tok/s | 2.26 s | $0.0002 | Agents & tools 48 · Coding 47 · Composite 38 · Human preference 56 · Instruction following 36 · Knowledge 33 · Long context 38 · Maths 52 · Reasoning 50 |
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| GPQA Diamond | 50.4%#199 | 70.1%#157 | 61.7%#176 | Epoch AI Benchmarking Hub |
| LMArena Hard Prompts | – | – | 1407#145 | LMArena |
| GPQA Diamond (AA)not in index | – | 70.7%#282 | 65.9%#322 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | – | 10.3%#252 | 6.9%#306 | Artificial Analysis |
| Kagi LLM Benchmark | – | 54.9%#67 | – | Kagi LLM Benchmark |
Coding
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| SciCode | – | 33.3%#137 | – | SciCode |
| WeirdML | – | – | 29.8%#127 | WeirdML |
| LMArena Coding | – | – | 1439#138 | LMArena |
| SciCode (AA)not in index | – | 33.0%#139 | – | Artificial Analysis |
Agents & tools
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| Terminal-Bench Hard | – | 5.3%#263 | 6.8%#240 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | – | 28.1%#261 | 22.2%#300 | Artificial Analysis |
| BFCL Overall | – | – | 41.4%#33 | Berkeley Function Calling Leaderboard |
Maths
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| OTIS Mock AIME | 25.6%#196 | 70.3%#124 | 62.8%#142 | Epoch AI Benchmarking Hub |
| AIME 2026 | – | 88.3%#29 | – | MathArena |
| HMMT February 2026 | – | 78.8%#27 | – | MathArena |
| MathArena Apex | – | 0.5%#41 | – | MathArena |
Knowledge
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| AA-Omniscience | – | -51.5#329 | -66.0#425 | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| IFBench | – | 50.7%#168 | 33.1%#319 | Artificial Analysis |
Human preference
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| LMArena Text | – | – | 1382#150 | LMArena |
Long context
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| AA-LCR | – | 61.3%#210 | 26.3%#335 | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | 139.7#100 | 137.4#109 | Epoch AI Benchmarking Hub |
| AA Intelligence Index | – | 9.8#319 | 7.5#390 | Artificial Analysis |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| DeepInfra | 112 tok/s | 0.96 s | $0.120 | $0.500 | 41k | fp8 |
| Alibaba Cloud Int. | 89 tok/s | 0.41 s | $0.130 | $0.520 | 131k | – |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.010 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0002 | $0.0002 | 5.0 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0017 | $0.0008 | 7.8 s |
| Code edit | 6,000 / 1,500 | $0.0015 | $0.0010 | 16.2 s |
| Agentic coding session | 60,000 / 4,000 | $0.0092 | $0.0043 | 39.4 s |
| Structured extraction | 2,000 / 200 | $0.0003 | $0.0002 | 4.1 s |
See also
Data as of 9 Sept 2026. Compare these configurations.