AlibabaQwen3-VL 235B-A22B
Best configuration ranks #256 of 610 on the BenchLeader Index at 52.1 ±3.9 (thinking reasoning effort). Last measured 2 Sept 2026. Released 4 Oct 2025.
- Blended price
- $1.30/M
- $0.400 in · $4.00 out
- Output speed
- 56 tok/s
- First answer
- 39 s
- first token 2.88 s
- Context
- 131k
- Overall index52
- Reasoning58
- Coding59
- Agents & tools43
- Knowledge42
- Instruction following57
- Human preference58
- Multimodal54
- Long context58
- Composite46
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 |
|---|---|---|---|---|---|---|
| thinkingbest | 52.1 | #256 | 56 tok/s | 39 s | $0.0014 | Agents & tools 43 · Coding 59 · Composite 46 · Human preference 58 · Instruction following 57 · Knowledge 42 · Long context 58 · Multimodal 54 · Reasoning 58 |
| default | 49.6 | #304 | 52 tok/s | 2.88 s | $0.0005 | Agents & tools 40 · Coding 60 · Composite 42 · Human preference 60 · Instruction following 45 · Knowledge 39 · Long context 42 · Multimodal 57 · Reasoning 61 |
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 | thinking | default | Source |
|---|---|---|---|
| LMArena Hard Prompts | 1418#133 | 1439#110 | LMArena |
| GPQA Diamond (AA)not in index | 77.2%#215 | 71.2%#278 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | 11.9%#224 | 6.6%#313 | Artificial Analysis |
Coding
| Benchmark | thinking | default | Source |
|---|---|---|---|
| LMArena Coding | 1453#124 | 1465#108 | LMArena |
Agents & tools
| Benchmark | thinking | default | Source |
|---|---|---|---|
| Terminal-Bench Hard | 11.4%#207 | 6.8%#240 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | 54.1%#182 | 35.1%#229 | Artificial Analysis |
Knowledge
| Benchmark | thinking | default | Source |
|---|---|---|---|
| AA-Omniscience | -46.5#292 | -52.7#338 | Artificial Analysis |
Instruction following
| Benchmark | thinking | default | Source |
|---|---|---|---|
| IFBench | 56.5%#140 | 42.6%#231 | Artificial Analysis |
Human preference
| Benchmark | thinking | default | Source |
|---|---|---|---|
| LMArena Text | 1395#137 | 1414#114 | LMArena |
Multimodal
| Benchmark | thinking | default | Source |
|---|---|---|---|
| LMArena Vision | 1206#71 | 1247#51 | LMArena |
| MMMU-Pro | 68.7%#130 | 67.6%#134 | Artificial Analysis |
| MMMU (validation) | – | 78.7%#8 | MMMU |
| MMMU-Pro (official)not in index | – | 68.1%#11 | MMMU |
Long context
| Benchmark | thinking | default | Source |
|---|---|---|---|
| AA-LCR | 63.7%#202 | 32.7%#310 | Artificial Analysis |
Composite
| Benchmark | thinking | default | Source |
|---|---|---|---|
| AA Intelligence Index | 13.4#243 | 9.9#314 | 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. | 25 tok/s | 0.69 s | $0.260 | $1.04 | 131k | – |
| Venice | 19 tok/s | 1.56 s | $0.210 | $1.90 | 128k | fp8 |
| Parasail | 11 tok/s | 1.09 s | $0.210 | $1.90 | 131k | fp8 |
| DeepInfra | 4 tok/s | 2.56 s | $0.200 | $0.880 | 262k | fp8 |
| NovitaAI | 0 tok/s | 13 s | $0.300 | $1.50 | 131k | bf16 |
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 $0.110 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0014 | $0.0013 | 44.2 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0072 | $0.0046 | 49.6 s |
| Code edit | 6,000 / 1,500 | $0.0084 | $0.0071 | 1.1 min |
| Agentic coding session | 60,000 / 4,000 | $0.040 | $0.027 | 1.8 min |
| Structured extraction | 2,000 / 200 | $0.0016 | $0.0012 | 42.4 s |
See also
Data as of 9 Sept 2026. Compare these configurations.