Alibabaqwen3-4b-instruct-2507
Best configuration ranks #478 of 610 on the BenchLeader Index at 41.7 ±3.8. Last measured 6 Aug 2025stale: no new result in six months. Released 6 Aug 2025.
- Blended price
- $0.030/M
- $0.030 in · $0.030 out
- Output speed
- –
- First answer
- –
- Context
- 128k
- Overall index42
- Reasoning40
- Agents & tools44
- Maths47
- Knowledge35
- Instruction following37
- Long context31
- Composite38
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 | – | – | – | – | $0.0000 | Reasoning 34 |
| thinking | 40.1 | #515 | – | – | $0.0000 | Agents & tools 35 · Composite 40 · Instruction following 51 · Knowledge 36 · Long context 44 · Maths 13 |
| defaultbest | 41.7 | #478 | – | – | $0.0000 | Agents & tools 44 · Composite 38 · Instruction following 37 · Knowledge 35 · Long context 31 · Maths 47 · Reasoning 40 |
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 | 43.6%#222 | – | 52.3%#195 | Epoch AI Benchmarking Hub |
| GPQA Diamond (AA)not in index | – | 66.7%#314 | 51.7%#395 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | – | 6.2%#331 | 4.5%#419 | Artificial Analysis |
Agents & tools
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| Terminal-Bench Hard | – | 1.5%#321 | 4.5%#270 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | – | 25.4%#280 | 26.6%#271 | Artificial Analysis |
| BFCL Overall | – | – | 35.7%#41 | Berkeley Function Calling Leaderboard |
Maths
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| OTIS Mock AIME | – | – | 52.2%#163 | Epoch AI Benchmarking Hub |
| AIME 2026 | – | 82.5%#30 | – | MathArena |
| HMMT February 2026 | – | 53.0%#31 | – | MathArena |
| MathArena Apex | – | 2.1%#29 | – | MathArena |
Knowledge
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| AA-Omniscience | – | -60.3#393 | -61.9#401 | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| IFBench | – | 49.8%#173 | 33.5%#311 | Artificial Analysis |
Long context
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| AA-LCR | – | 37.3%#297 | 11.3%#390 | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | default | Source |
|---|---|---|---|---|
| AA Intelligence Index | – | 8.8#349 | 6.7#426 | Artificial Analysis |
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 | $0.0000 | – | – |
| Summarise a 30-page report | 12,000 / 600 | $0.0004 | – | – |
| Code edit | 6,000 / 1,500 | $0.0002 | – | – |
| Agentic coding session | 60,000 / 4,000 | $0.0019 | – | – |
| Structured extraction | 2,000 / 200 | $0.0001 | – | – |
See also
Data as of 9 Sept 2026. Compare these configurations.