AlibabaQwen3.5 397B A17B
Qwen3.5 397B A17B is an Alibaba open-weights model, released 13 Feb 2026. Its best configuration (thinking reasoning effort) ranks #235 of 785 on the BenchLeader Index at 55.1 ±7.6, in the upper half. The ± is the point: 376 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 (77) and lowest in reasoning (44). At $1.35 per million tokens blended it is pricier than most ranked models. Output speed of 57 tokens per second puts it slower than most, with a first answer in 58.2 s. It has been measured at 3 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 2 Mar 2026.stale: no new result in six months
- Blended price
- $1.35/M
- $0.600 in · $3.60 out
- Output speed
- 57 tok/s
- measured by Artificial Analysis
- First answer
- 58 s
- first token 0.59 s
- Context
- 262k
- Full answer
- 67 s
- median, reasoning included
- Cost per run
- $0.475
- one full Intelligence Index run
- Released
- 13 Feb 2026
- Overall index55
- Reasoning44
- Agents & tools56
- Knowledge48
- Instruction following77
- Multimodal60
- Long context64
- Composite50
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 | 50.8 | #359 | 58 tok/s | 2.13 s | $0.0013 | Agents & tools 65 · Composite 53 · Instruction following 53 · Knowledge 44 · Long context 57 · Maths 48 · Multimodal 35 · Reasoning 52 |
| thinkingbest | 55.1 | #235 | 57 tok/s | 58 s | $0.0013 | Agents & tools 56 · Composite 50 · Instruction following 77 · Knowledge 48 · Long context 64 · Multimodal 60 · Reasoning 44 |
| not stated | 51.5 | #342 | 50 tok/s | 0.59 s | $0.0004 | Agents & tools 24 · Coding 50 · Human preference 63 · Maths 52 · Multimodal 60 · Reasoning 63 |
How its index has moved
8 Sept 2026 to 23 Oct 2026- thinking
- not stated
- no reasoning
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
See all LLM benchmarks. 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 | 86.4%#78 | – | 85.9%#84 | Epoch AI Benchmarking Hub | |
| Arena Hard Promptseffort not stated | – | – | 1464#88 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 86.1%#114 | 89.3%#79 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 19.8%#216 | 29.0%#159 | – | Artificial Analysis | |
| Kagi LLM Benchmarkeffort not stated | – | – | 73.7%#16 | Kagi LLM Benchmark | |
| CritPt | 0.9%#237 | 1.7%#206 | – | Artificial Analysis | |
| Mystery Game Puzzlesnot in index | 18.0%#69 | – | – | Epoch AI Benchmarking Hub | |
| LMCAnot in indexeffort not stated | – | – | 37.9%#115 | Epoch AI Benchmarking Hub | |
| DTBenchnot in indexeffort not stated | – | – | 87.5%#100 | Epoch AI Benchmarking Hub |
Coding
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Codingeffort not stated | – | 1492#92 | Arena (LMArena) | |
| Arena WebDeveffort not stated | – | 1400#89 | Arena (LMArena) | |
| SciCode (AA)not in index | 44.8%#149 | – | Artificial Analysis | |
| Terminal-Bench 4.0 (AA)not in index | 0.0%#152 | – | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | 51.3%#117 | – | Artificial Analysis |
Agents & tools
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| APEX-Agentseffort not stated | – | – | 24.9%#51 | Mercor | |
| Terminal-Bench Hard | 35.6%#74 | 40.9%#53 | – | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 83.9%#105 | 95.6%#19 | – | Artificial Analysis | |
| τ²-bench | – | 87.9%#1 | – | τ²-bench | |
| τ³-bench Bankingnot in index | – | 9.8%#25 | – | τ²-bench | |
| GDPval-AA v2.1not in index | – | 14.6%#105 | – | Artificial Analysis | |
| τ³-Banking (AA) | – | 13.4%#126 | – | Artificial Analysis | |
| ITBench SRE (AA)not in index | – | 34.1%#28 | – | Artificial Analysis | |
| APEX-Agents (AA)not in index | – | 15.3%#22 | – | Artificial Analysis | |
| AutomationBenchnot in index | – | 6.9%#157 | – | Artificial Analysis | |
| GDP.pdf (AA)not in index | – | 12.2%#114 | – | Artificial Analysis | |
| EnterpriseOps-Gymnot in index | – | 30.9%#40 | – | Artificial Analysis |
Maths
| Benchmark | no reasoning | not stated | Source | Trend |
|---|---|---|---|---|
| FrontierMath Tiers 1–3 | 31.2%#73 | 29.5%#75 | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | 82.2%#108 | 88.9%#76 | Epoch AI Benchmarking Hub | |
| AIME 2026effort not statedpossible contamination | – | 94.2%#20 | MathArena | |
| HMMT February 2026effort not statedpossible contamination | – | 87.9%#17 | MathArena | |
| USAMO 2026not in indexeffort not stated | – | 36.3%#8 | MathArena | |
| Arena Mathseffort not stated | – | 1451#75 | Arena (LMArena) |
Knowledge
| Benchmark | no reasoning | thinking | Source | Trend |
|---|---|---|---|---|
| AA-Omniscience | -37.9#317 | -30.8#284 | Artificial Analysis | |
| AA-Omniscience: accuracynot in index | 24.5%#241 | 30.8%#178 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 17.3%#318 | 11.1%#396 | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| IFBench | 51.6%#166 | 78.8%#13 | – | Artificial Analysis | |
| Arena Instruction Followingnot in indexeffort not stated | – | – | 1434#90 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1442#90 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1407#97 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1453#81 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1453#82 | Arena (LMArena) |
Multimodal
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Arena Visioneffort not stated | – | – | 1263#54 | Arena (LMArena) | |
| MMMU-Pro | 52.7%#236 | 77.3%#83 | – | Artificial Analysis |
Long context
| Benchmark | no reasoning | thinking | Source | Trend |
|---|---|---|---|---|
| AA-LCR | 64.3%#257 | 77.3%#132 | Artificial Analysis | |
| MLCRnot in index | – | 8.9%#122 | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | 146.8#71 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 21.4#214 | 18.4#254 | – | 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 |
|---|---|---|---|---|---|---|
| StreamLake | 94 tok/s | 0.94 s | $0.600 | $3.60 | 256k | – |
| Alibaba Cloud Int. | 52 tok/s | 1.37 s | $0.390 | $2.34 | 262k | – |
| AtlasCloud | 51 tok/s | 1.08 s | $0.550 | $3.50 | 262k | fp8 |
| Parasail | 48 tok/s | 0.59 s | $0.500 | $3.60 | 262k | fp8 |
| Venice | 45 tok/s | 1.25 s | $0.750 | $4.50 | 128k | – |
| NovitaAI | 42 tok/s | 2.74 s | $0.600 | $3.60 | 262k | – |
| DeepInfra | 41 tok/s | 0.36 s | $0.450 | $3.00 | 262k | fp8 |
| Phala | 28 tok/s | 1.13 s | $0.550 | $3.50 | 262k | – |
Price history
The price above, per 1M tokens, every day it changed.
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 $0.330 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0013 | $0.0012 | 1.1 min |
| Summarize a 30-page report | 12,000 / 600 | $0.0094 | $0.0069 | 1.1 min |
| Code edit | 6,000 / 1,500 | $0.0090 | $0.0078 | 1.4 min |
| Agentic coding session | 60,000 / 4,000 | $0.050 | $0.038 | 2.1 min |
| Structured extraction | 2,000 / 200 | $0.0019 | $0.0015 | 1.0 min |
More from Alibaba
See also
Questions people ask
- What are Qwen3.5 397B A17B's benchmark scores?
- As of 24 Oct 2026, Qwen3.5 397B A17B (thinking effort) scores 55.1 ±7.6 on the BenchLeader Index, strongest in instruction following (77) and weakest in reasoning (44). That rests on 9 independent benchmark results, and every score on this page links to its publisher.
- How does Qwen3.5 397B A17B rank?
- Qwen3.5 397B A17B (thinking effort) is #235 of 785 ranked configurations on 24 Oct 2026. 376 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Qwen3.5 397B A17B cost per 1M tokens?
- As of 24 Oct 2026, Qwen3.5 397B A17B costs $0.600 per 1M input tokens and $3.60 per 1M output tokens, $1.35 blended (three input tokens per output token). The list price last changed on 23 Sept 2026.
- When was Qwen3.5 397B A17B released?
- Alibaba released Qwen3.5 397B A17B on 13 Feb 2026. BenchLeader has results for it at 3 reasoning-effort settings.
Data as of 24 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Qwen3.5 397B A17B: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/qwen3-5-397b-a17b.