AlibabaQwen3 235B A22B
Best configuration ranks #267 of 610 on the BenchLeader Index at 51.7 ±3.6 (thinking reasoning effort). Last measured 2 Sept 2026. Released 28 Apr 2025.
- Blended price
- $0.975/M
- $0.300 in · $3.00 out
- Output speed
- 60 tok/s
- First answer
- 37 s
- first token 2.37 s
- Context
- 256k
- Overall index52
- Reasoning60
- Coding50
- Agents & tools45
- Maths52
- Knowledge46
- Instruction following52
- Human preference58
- Long context60
- Composite45
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 | – | – | 58 tok/s | 2.37 s | $0.0005 | Reasoning 51 |
| thinkingbest | 51.7 | #267 | 60 tok/s | 37 s | $0.0010 | Agents & tools 45 · Coding 50 · Composite 45 · Human preference 58 · Instruction following 52 · Knowledge 46 · Long context 60 · Maths 52 · Reasoning 60 |
| default | 48.7 | #317 | 62 tok/s | 2.67 s | $0.0005 | Agents & tools 57 · Coding 48 · Composite 44 · Human preference 61 · Instruction following 38 · Knowledge 48 · Long context 42 · Maths 57 · Reasoning 42 |
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 | – | 80.0%#113 | 70.7%#156 | Epoch AI Benchmarking Hub |
| SimpleBench | – | – | 31.0%#66 | SimpleBench |
| LMArena Hard Prompts | – | 1417#136 | 1448#94 | LMArena |
| GPQA Diamond (AA)not in index | – | 79.0%#197 | 75.3%#240 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | – | 15.9%#189 | 11.1%#235 | Artificial Analysis |
| GPQA Diamond (Vals)not in index | – | – | 70.2%#89 | Vals AI |
| Kagi LLM Benchmark | 55.0%#66 | 69.4%#29 | – | Kagi LLM Benchmark |
| ARC-AGI-1 | – | – | 11.0%#168 | ARC Prize |
| ARC-AGI-2 | – | – | 1.3%#153 | ARC Prize |
Coding
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| SciCode | – | 42.4%#97 | – | SciCode |
| WeirdML | – | 41.0%#96 | 38.7%#109 | WeirdML |
| LMArena Coding | – | 1443#134 | 1472#100 | LMArena |
| SciCode (AA)not in index | – | 41.4%#109 | – | Artificial Analysis |
| LiveCodeBench | – | – | 70.6%#84 | Vals AI |
| IOI | – | – | 0.0%#58 | Vals AI |
| SWE-Bench Pro | – | – | 21.4%#18 | Scale AI SEAL |
| Aider Polyglot | – | – | 59.6%#16 | Aider polyglot leaderboard |
Agents & tools
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| Terminal-Bench Hard | – | 13.6%#194 | 15.2%#186 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | – | 53.2%#184 | 33.3%#237 | Artificial Analysis |
| BFCL Overall | – | – | 52.1%#21 | Berkeley Function Calling Leaderboard |
Maths
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| OTIS Mock AIME | – | 86.7%#71 | – | Epoch AI Benchmarking Hub |
| MATH Level 5 | – | – | 68.9%#39 | Epoch AI Benchmarking Hub |
| AIME (Vals) | – | – | 84.0%#40 | Vals AI |
| MGSM | – | – | 92.5%#17 | Vals AI |
| MathArena Apex | – | 5.2%#25 | – | MathArena |
Knowledge
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| SimpleQA Verified | – | 40.4%#38 | – | Epoch AI Benchmarking Hub |
| AA-Omniscience | – | -44.7#283 | -44.0#279 | Artificial Analysis |
| MMLU-Pro | – | – | 81.3%#79 | Vals AI |
| LegalBench | – | – | 80.2%#82 | Vals AI |
| TaxEval | – | – | 70.7%#86 | Vals AI |
| MedQA | – | – | 90.6%#44 | Vals AI |
| MultiNRC | – | 27.1%#28 | 17.6%#36 | Scale AI SEAL |
Instruction following
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| IFBench | – | 51.2%#165 | 46.0%#194 | Artificial Analysis |
| MultiChallenge | – | – | 41.2%#26 | Scale AI SEAL |
Human preference
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| LMArena Text | – | 1400#132 | 1423#104 | LMArena |
Long context
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| Fiction.LiveBench 120k | – | 68.8%#8 | – | Fiction.live |
| AA-LCR | – | 72.0%#140 | 33.9%#308 | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | 143.9#78 | 139.4#103 | Epoch AI Benchmarking Hub |
| AA Intelligence Index | – | 12.7#258 | 12#272 | 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. | 31 tok/s | 0.50 s | $0.455 | $1.82 | 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. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0010 | – | 41.8 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0054 | – | 46.8 s |
| Code edit | 6,000 / 1,500 | $0.0063 | – | 1.0 min |
| Agentic coding session | 60,000 / 4,000 | $0.030 | – | 1.7 min |
| Structured extraction | 2,000 / 200 | $0.0012 | – | 40.2 s |
See also
Data as of 9 Sept 2026. Compare these configurations.