AlibabaQwen3.5-9B
Qwen3.5-9B is an Alibaba open-weights reasoning model, released 24 Feb 2026. Its best configuration (thinking reasoning effort) ranks #387 of 751 on the BenchLeader Index at 49.1 ±7.3, in the lower half. The ± is the point: 400 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in instruction following (66) and lowest in knowledge (38). At $0.113 per million tokens blended it is among the cheapest fifth of ranked models. Output speed of 52 tokens per second puts it slower than most, with a first answer in 40.8 s. It has been measured at 3 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them.
- Blended price
- $0.113/M
- $0.100 in · $0.150 out
- Output speed
- 52 tok/s
- measured by Artificial Analysis
- First answer
- 41 s
- first token 0.74 s
- Context
- 262k
- Full answer
- 50 s
- median, reasoning included
- Cost per run
- $0.213
- one full Intelligence Index run
- Released
- 24 Feb 2026
- Overall index49
- Reasoning42
- Agents & tools41
- Knowledge38
- Instruction following66
- Multimodal52
- Long context60
- Composite42
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. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoning | 46.5 | #461 | 96 tok/s | 0.66 s | $0.0001 | Agents & tools 49 · Composite 45 · Instruction following 41 · Knowledge 30 · Long context 48 · Maths 51 · Multimodal 50 · Reasoning 50 |
| thinkingbest | 49.1 | #387 | 52 tok/s | 41 s | $0.0001 | Agents & tools 41 · Composite 42 · Instruction following 66 · Knowledge 38 · Long context 60 · Multimodal 52 · Reasoning 42 |
| not stated | 40.1 | #636 | 33 tok/s | 0.74 s | $0.0001 | Agents & tools 24 · Coding 25 · Maths 37 · Reasoning 58 |
How its index has moved
9 Sept 2026 to 3 Oct 2026- thinking
- no reasoning
- not stated
The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
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 | not stated | Source | Trend |
|---|---|---|---|---|---|
| GPQA Diamond | 78.9%#125 | – | 79.0%#124 | Epoch AI Benchmarking Hub | |
| GPQA Diamond (AA)not in index | 78.6%#204 | 80.6%#186 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 9.4%#315 | 14.9%#235 | – | Artificial Analysis | |
| CritPt | 0.6%#234 | 0.3%#250 | – | Artificial Analysis | |
| Chess Puzzlesnot in index | 12.0%#123 | – | 2.0%#178 | Epoch AI Benchmarking Hub | |
| LMCAnot in index | – | – | 24.5%#129 | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | – | – | 71.2%#134 | Epoch AI Benchmarking Hub |
Coding
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| SciCode | – | – | 27.6%#176 | SciCode | |
| SciCode (AA)not in index | – | 29.5%#193 | – | Artificial Analysis | |
| Terminal-Bench 4.0 (AA)not in index | – | 0.5%#123 | – | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | 21.4%#166 | 29.2%#153 | – | Artificial Analysis |
Agents & tools
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Terminal-Bench | – | – | 9.2%#74 | Terminal-Bench | |
| Terminal-Bench Hard | 18.2%#170 | 24.2%#142 | – | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 85.1%#93 | 86.8%#83 | – | Artificial Analysis | |
| GDPval (AA)not in index | – | 0.0%#197 | – | Artificial Analysis | |
| τ²-Bench Banking (AA) | – | 7.0%#160 | – | Artificial Analysis |
Maths
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| OTIS Mock AIME | 61.7%#154 | – | 42.8%#187 | 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 | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| AA-Omniscience | -70.7#491 | -53.5#395 | – | Artificial Analysis | |
| AA-Omniscience: accuracynot in index | 14.0%#420 | 16.4%#365 | – | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 1.6%#525 | 16.4%#310 | – | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| IFBench | 37.8%#283 | 66.7%#90 | – | Artificial Analysis |
Multimodal
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| MMMU-Pro | 66.8%#163 | 69.3%#148 | – | Artificial Analysis |
Long context
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| AA-LCR | 46.0%#310 | 70.0%#199 | – | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | – | 139.5#107 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index | 13.3#292 | 11.2#337 | – | 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 |
|---|---|---|---|---|---|---|
| Venice | 122 tok/s | 1.06 s | $0.100 | $0.150 | 256k | fp8 |
| Parasail | 89 tok/s | 0.63 s | $0.100 | $0.250 | 262k | bf16 |
| Together | 68 tok/s | 1.55 s | $0.170 | $0.250 | 262k | – |
| Darkbloom | 47 tok/s | 1.50 s | $0.080 | $0.130 | 262k | fp4 |
| SiliconFlow | 33 tok/s | 1.15 s | $0.100 | $0.150 | 262k | fp8 |
| DeepInfra | 13 tok/s | 1.29 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 | – | 46.6 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0013 | – | 52.4 s |
| Code edit | 6,000 / 1,500 | $0.0008 | – | 1.2 min |
| Agentic coding session | 60,000 / 4,000 | $0.0066 | – | 2.0 min |
| Structured extraction | 2,000 / 200 | $0.0002 | – | 44.7 s |
See also
Data as of 4 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Qwen3.5-9B: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/qwen3-5-9b, data as of 4 Oct 2026.