AlibabaQwen3.5-9B
Best configuration ranks #347 of 610 on the BenchLeader Index at 47.7 ±5.3 (no reasoning reasoning effort). Last measured 24 Feb 2026stale: no new result in six months. Released 24 Feb 2026.
- Blended price
- $0.190/M
- $0.170 in · $0.250 out
- Output speed
- 89 tok/s
- First answer
- 0.74 s
- first token 0.74 s
- Context
- 262k
- Overall index48
- Reasoning58
- Agents & tools49
- Maths51
- Knowledge31
- Instruction following41
- Multimodal51
- Long context49
- 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 |
|---|---|---|---|---|---|---|
| no reasoningbest | 47.7 | #347 | 89 tok/s | 0.74 s | $0.0001 | Agents & tools 49 · Composite 46 · Instruction following 41 · Knowledge 31 · Long context 49 · Maths 51 · Multimodal 51 · Reasoning 58 |
| default | 47.5 | #351 | 83 tok/s | 26 s | $0.0001 | Agents & tools 35 · Coding 27 · Composite 46 · Instruction following 65 · Knowledge 39 · Long context 61 · Maths 37 · Multimodal 54 · Reasoning 58 |
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 | default | Source |
|---|---|---|---|
| GPQA Diamond | 78.9%#119 | 79.0%#118 | Epoch AI Benchmarking Hub |
| GPQA Diamond (AA)not in index | 78.6%#202 | 80.6%#185 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | 9.4%#269 | 14.9%#195 | Artificial Analysis |
Agents & tools
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| Terminal-Bench | – | 9.2%#65 | Terminal-Bench |
| Terminal-Bench Hard | 18.2%#166 | 24.2%#137 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | 85.1%#92 | 86.8%#82 | Artificial Analysis |
Maths
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| OTIS Mock AIME | 61.7%#145 | 42.8%#178 | Epoch AI Benchmarking Hub |
| AIME 2026 | – | 92.5%#23 | MathArena |
| HMMT February 2026 | – | 71.2%#29 | MathArena |
| MathArena Apex | – | 0.5%#41 | MathArena |
Knowledge
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| AA-Omniscience | -70.7#439 | -53.5#346 | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| IFBench | 37.8%#275 | 66.7%#90 | Artificial Analysis |
Multimodal
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| MMMU-Pro | 66.8%#136 | 69.3%#121 | Artificial Analysis |
Long context
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| AA-LCR | 46.0%#264 | 70.0%#161 | Artificial Analysis |
Composite
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | 139.4#101 | Epoch AI Benchmarking Hub |
| AA Intelligence Index | 13.3#247 | 13.7#239 | 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 |
|---|---|---|---|---|---|---|
| Together | 49 tok/s | 0.47 s | $0.170 | $0.250 | 262k | – |
| Venice | 42 tok/s | 1.08 s | $0.100 | $0.150 | 256k | fp8 |
| Parasail | 26 tok/s | 4.78 s | $0.100 | $0.250 | 262k | bf16 |
| Darkbloom | 23 tok/s | 1.35 s | $0.080 | $0.130 | 262k | fp4 |
| SiliconFlow | 22 tok/s | 1.64 s | $0.100 | $0.150 | 262k | fp8 |
| DeepInfra | 15 tok/s | 1.15 s | $0.100 | $0.150 | 262k | bf16 |
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 | $0.0001 | – | 4.1 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0022 | – | 7.5 s |
| Code edit | 6,000 / 1,500 | $0.0014 | – | 17.6 s |
| Agentic coding session | 60,000 / 4,000 | $0.011 | – | 45.7 s |
| Structured extraction | 2,000 / 200 | $0.0004 | – | 3.0 s |
See also
Data as of 9 Sept 2026. Compare these configurations.