AlibabaQwen3.5 35B-A3B
Best configuration ranks #223 of 610 on the BenchLeader Index at 53.4 ±5.2. Last measured 8 Sept 2026. Released 24 Feb 2026.
- Blended price
- $0.688/M
- $0.250 in · $2.00 out
- Output speed
- 145 tok/s
- First answer
- 16 s
- first token 1.34 s
- Context
- 262k
- Overall index53
- Reasoning60
- Coding39
- Agents & tools57
- Maths44
- Knowledge41
- Instruction following70
- Human preference58
- Multimodal57
- Long context62
- Composite53
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 | 48.0 | #337 | 164 tok/s | 2.10 s | $0.0007 | Agents & tools 43 · Coding 29 · Composite 48 · Instruction following 46 · Knowledge 34 · Long context 58 · Maths 55 · Multimodal 54 · Reasoning 59 |
| defaultbest | 53.4 | #223 | 145 tok/s | 16 s | $0.0007 | Agents & tools 57 · Coding 39 · Composite 53 · Human preference 58 · Instruction following 70 · Knowledge 41 · Long context 62 · Maths 44 · Multimodal 57 · Reasoning 60 |
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 | Trend |
|---|---|---|---|---|
| GPQA Diamond | 81.2%#109 | 83.5%#89 | Epoch AI Benchmarking Hub | |
| LMArena Hard Prompts | – | 1413#143 | LMArena | |
| GPQA Diamond (AA)not in index | 81.9%#171 | 84.5%#136 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 13.4%#208 | 21.0%#156 | Artificial Analysis |
Coding
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| SciCode | 29.3%#141 | – | SciCode | |
| LMArena Coding | – | 1435#145 | LMArena | |
| LMArena WebDev | – | 1250#106 | LMArena |
Agents & tools
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| Terminal-Bench Hard | 10.6%#213 | 26.5%#126 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 86.3%#87 | 89.2%#70 | Artificial Analysis |
Maths
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| OTIS Mock AIME | 70.0%#125 | 54.4%#160 | Epoch AI Benchmarking Hub | |
| AIME 2026 | – | 93.3%#22 | MathArena | |
| HMMT February 2026 | – | 81.8%#25 | MathArena | |
| MathArena Apex | – | 4.2%#28 | MathArena |
Knowledge
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| AA-Omniscience | -63.0#409 | -48.1#300 | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| IFBench | 44.5%#207 | 72.5%#47 | Artificial Analysis |
Human preference
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| LMArena Text | – | 1395#136 | LMArena |
Multimodal
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| MMMU-Pro | 69.2%#124 | 72.7%#102 | Artificial Analysis |
Long context
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| AA-LCR | 63.0%#206 | 72.0%#140 | Artificial Analysis |
Composite
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | 142.5#85 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index | 15.1#222 | 19.3#183 | 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. | 147 tok/s | 0.75 s | $0.163 | $1.30 | 262k | – |
| AtlasCloud | 114 tok/s | 1.34 s | $0.225 | $1.80 | 262k | fp8 |
| Venice | 61 tok/s | 1.57 s | $0.313 | $1.25 | 256k | – |
| Parasail | 58 tok/s | 0.51 s | $0.150 | $1.00 | 262k | fp8 |
| DeepInfra | 42 tok/s | 0.43 s | $0.140 | $1.00 | 262k | fp8 |
| SiliconFlow | 28 tok/s | 1.78 s | $0.240 | $1.80 | 262k | fp8 |
| Darkbloom | 17 tok/s | 2.36 s | $0.080 | $0.750 | 262k | fp4 |
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 at $0.050 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0007 | $0.0006 | 18.0 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0042 | $0.0024 | 20.1 s |
| Code edit | 6,000 / 1,500 | $0.0045 | $0.0036 | 26.3 s |
| Agentic coding session | 60,000 / 4,000 | $0.023 | $0.014 | 43.6 s |
| Structured extraction | 2,000 / 200 | $0.0009 | $0.0006 | 17.3 s |
See also
Data as of 9 Sept 2026. Compare these configurations.