AlibabaQwen3 Max
Best configuration ranks #186 of 610 on the BenchLeader Index at 55.0 ±6.7 (thinking reasoning effort). Released 5 Sept 2025.
- Blended price
- $2.40/M
- $1.20 in · $6.00 out
- Output speed
- 58 tok/s
- First answer
- 38 s
- first token 4.28 s
- Context
- 262k
- Overall index55
- Agents & tools55
- Knowledge47
- Instruction following69
- Long context64
- Composite56
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 |
|---|---|---|---|---|---|---|
| max | 52.7 | #240 | 48 tok/s | 2.38 s | $0.0014 | Agents & tools 51 · Coding 55 · Composite 49 · Human preference 61 · Instruction following 46 · Knowledge 55 · Long context 51 · Maths 52 · Reasoning 59 |
| thinkingbest | 55.0 | #186 | 58 tok/s | 38 s | $0.0023 | Agents & tools 55 · Composite 56 · Instruction following 69 · Knowledge 47 · Long context 64 |
| default | 51.8 | #264 | 57 tok/s | 4.28 s | $0.0023 | Agents & tools 51 · Coding 49 · Composite 45 · Human preference 63 · Instruction following 49 · Knowledge 50 · Long context 47 · Maths 52 · Reasoning 63 |
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 | max | thinking | default | Source |
|---|---|---|---|---|
| GPQA Diamond | 72.6%#148 | – | – | Epoch AI Benchmarking Hub |
| LMArena Hard Prompts | 1447#97 | – | 1459#82 | LMArena |
| GPQA Diamond (AA)not in index | 76.4%#226 | 86.1%#113 | 76.4%#226 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | 11.9%#224 | 28.0%#120 | 10.2%#255 | Artificial Analysis |
| GPQA Diamond (Vals)not in index | 84.8%#48 | – | 77.8%#73 | Vals AI |
| Kagi LLM Benchmark | 72.5%#22 | – | – | Kagi LLM Benchmark |
Coding
| Benchmark | max | thinking | default | Source |
|---|---|---|---|---|
| LMArena Coding | 1474#98 | – | 1481#88 | LMArena |
| LiveCodeBench | 78.2%#74 | – | 66.9%#94 | Vals AI |
| IOI | 15.7%#28 | – | 7.8%#35 | Vals AI |
Agents & tools
| Benchmark | max | thinking | default | Source |
|---|---|---|---|---|
| Terminal-Bench Hard | 20.4%#157 | 24.2%#137 | 19.7%#160 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | 74.3%#136 | 83.6%#105 | 32.8%#239 | Artificial Analysis |
Maths
| Benchmark | max | thinking | default | Source |
|---|---|---|---|---|
| FrontierMath Tiers 1–3 | 18.9%#78 | – | – | Epoch AI Benchmarking Hub |
| OTIS Mock AIME | 73.3%#114 | – | – | Epoch AI Benchmarking Hub |
| MATH Level 5 | 97.1%#7 | – | – | Epoch AI Benchmarking Hub |
| AIME (Vals) | 81.0%#46 | – | 60.7%#56 | Vals AI |
| MGSM | 91.8%#25 | – | 92.2%#21 | Vals AI |
Knowledge
| Benchmark | max | thinking | default | Source |
|---|---|---|---|---|
| SimpleQA Verified | 48.8%#25 | – | – | Epoch AI Benchmarking Hub |
| AA-Omniscience | -43.5#277 | -36.6#241 | -42.8#273 | Artificial Analysis |
| MMLU-Pro | 85.0%#58 | – | 83.5%#70 | Vals AI |
| LegalBench | 81.9%#72 | – | 80.3%#80 | Vals AI |
| CorpFin | 68.0%#11 | – | 53.0%#96 | Vals AI |
| TaxEval | 73.5%#47 | – | 74.0%#42 | Vals AI |
| MedQA | 87.4%#56 | – | 87.4%#55 | Vals AI |
Instruction following
| Benchmark | max | thinking | default | Source |
|---|---|---|---|---|
| IFBench | 44.1%#212 | 70.8%#64 | 48.0%#183 | Artificial Analysis |
Human preference
| Benchmark | max | thinking | default | Source |
|---|---|---|---|---|
| LMArena Text | 1424#103 | – | 1435#87 | LMArena |
Long context
| Benchmark | max | thinking | default | Source |
|---|---|---|---|---|
| AA-LCR | 50.0%#253 | 74.3%#116 | 43.0%#277 | Artificial Analysis |
Composite
| Benchmark | max | thinking | default | Source |
|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | 142.4#87 | – | – | Epoch AI Benchmarking Hub |
| AA Intelligence Index | 15.6#213 | 21.3#161 | 12.6#260 | 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 |
|---|---|---|---|---|---|---|
| Alibaba Cloud Int. | 33 tok/s | 1.11 s | $0.780 | $3.90 | 262k | – |
Price history
Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.
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. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0023 | – | 43.7 s |
| Summarise a 30-page report | 12,000 / 600 | $0.018 | – | 48.8 s |
| Code edit | 6,000 / 1,500 | $0.016 | – | 1.1 min |
| Agentic coding session | 60,000 / 4,000 | $0.096 | – | 1.8 min |
| Structured extraction | 2,000 / 200 | $0.0036 | – | 41.9 s |
See also
Data as of 9 Sept 2026. Compare these configurations.