NEWHow AI models draw the world map · Gecko Tests
Beta
Home/Models/Qwen3.5 397B A17B
Alibaba Qwen logo

Qwen3.5 397B A17B

by Alibaba Qwen · Released Feb 2026

Open SourceMultimodal
69.4
avg score
Rank #41
Compare
Better than 87% of all models
Context
262K tokens (~131 books)
Input $/1M
$0.55
Output $/1M
$3.50
Type
multimodal
License
Open Source
Benchmarks
29 tested
Data as of
About

The Qwen3.5 series 397B-A17B native vision-language model is built on a hybrid architecture that integrates a linear attention mechanism with a sparse mixture-of-experts model, achieving higher inference efficiency. It delivers...

Tested on 29 benchmarks · BenchGecko score 69.4. Top scores: Chatbot Arena Elo — Overall (1441.8%), Chatbot Arena Elo — Coding (1399.4%), OpenCompass — AIME2025 (92.3%).

Looking for similar performance at lower cost?
DeepSeek V4 Flash scores 68.6 (99% as good) at $0.09/1M input · 83% cheaper
Capabilities
coding
83.0
#4 globally
math
70.8
#46 globally
knowledge
58.8
#69 globally
agentic
24.9
#53 globally
speed
41.8
#61 globally
general
44.5
#55 globally
language
91.5
#9 globally
Benchmark Scores
Compare All
Tested on 29 benchmarks · Ranked across 8 categories
Score Distribution (all 319 models)
0255075100
▲ You are here
codingCompare coding →
OpenCompass — LiveCodeBenchV6

OpenCompass Live Code Bench v6. Fresh competitive programming problems to evaluate code generation without memorization.

83.0·
mathCompare math →
OpenCompass — AIME2025

OpenCompass evaluation on AIME 2025 problems. Tests mathematical reasoning on fresh competition problems.

92.3·
OTIS Mock AIME 2024-2025

Mock AIME (American Invitational Mathematics Exam) problems from OTIS. Tests mathematical competition performance.

88.9·
FrontierMath-Tiers-1-3-v2-Private
31.2·
knowledgeCompare knowledge →
OpenCompass — GPQA-Diamond

OpenCompass evaluation of GPQA Diamond. PhD-level science questions from the hardest subset.

88.4·
OpenCompass — MMLU-Pro

OpenCompass MMLU-Pro evaluation. Harder knowledge test with more answer choices.

87.6·
GPQA diamond

Graduate-level science questions written by PhD experts. Diamond subset contains questions where experts disagree, testing deep understanding.

81.8·
Quick compare:Microsoft logovs phi-3-medium 14BGoogle DeepMind logovs Gemini 3.8 FlashAnthropic logovs Claude Instant
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Model Family · Alibaba Qwen Qwen 3.5
Feb 2026
69.4
$0.55/M in262Kctx29 benchmarks
Feb 2026
$0.26/M in(-0.29)1.0Mctx(+738K)
Apr 2026
$0.30/M in(+0.04)1.0Mctx
Feb 2026
$0.26/M in(-0.04)262Kctx(-738K)15 benchmarks
Feb 2026
$0.20/M in(-0.07)262Kctx5 benchmarks
Feb 2026
53.3-16.1
$0.15/M in(-0.05)262Kctx10 benchmarks
Mar 2026
39.8-13.5
$0.10/M in(-0.05)262Kctx19 benchmarks
Mar 2026
$0.17/M in(+0.07)262Kctx
Feb 2026
35.3-4.5
$0.07/M in(-0.11)1.0Mctx(+738K)12 benchmarks
See the full Qwen 3.5 family →
Recently Happened
Qwen3.5 397B A17B pricing increased 50%
Sep 9, 2026
Qwen3.5 397B A17B pricing dropped 33%
Sep 7, 2026
Qwen3.5 397B A17B pricing increased 50%
Sep 3, 2026
Qwen3.5 397B A17B pricing dropped 35%
Aug 26, 2026
Qwen3.5 397B A17B pricing increased 54%
Aug 24, 2026
Qwen3.5 397B A17B pricing dropped 35%
Aug 15, 2026
Similar Models
Microsoft logo
phi-3-medium 14B
Microsoft
69.2n/a
Google DeepMind logo
Gemini 3.8 Flash
Google DeepMind
68.8$0.82/1M
Anthropic logo
Claude Instant
Anthropic
68.7n/a
Links
Info
Alibaba QwenPricing explorerDevelopers · API
Research
Technical Report
Documentation
API DocsPlayground
Community
@Alibaba Qwen
Source Code
GitHubHugging Face
BenchGecko API
qwen3-5-397b-a17b
Specifications
  • Typemultimodal
  • Context262K tokens (~131 books)
  • ReleasedFeb 2026
  • LicenseOpen Source
  • StatusActive
  • Cost / Message~$0.005
Available On
Alibaba Qwen logoAlibaba Qwen$0.55
Categories
codingmathknowledgeagenticspeedgenerallanguage
Learn More
context-windowtransformeropen weightstokens
Share & Export
Tweet
Related Models
Microsoft logophi-3-medium 14B
69.2
Google DeepMind logoGemini 3.8 Flash
68.8
Anthropic logoClaude Instant
68.7
Alibaba Qwen logoQwen3.7 Max
70.2
Qwen3.5 397B A17B is an open-source multimodal AI model by Alibaba Qwen, released in February 2026. It has an average benchmark score of 69.4. Context window: 262K tokens.

Related Models

phi-3-medium 14B · MicrosoftGemini 3.8 Flash · Google DeepMindClaude Instant · AnthropicQwen3.7 Max · Alibaba QwenDeepSeek V4 Flash · DeepSeek

Benchmarks

Chatbot Arena Elo — OverallChatbot Arena Elo — CodingOpenCompass — AIME2025OpenCompass — IFEvalOTIS Mock AIME 2024-2025

Related Pages

Alibaba Qwen · ProviderAlibaba Qwen · EconomyAll ModelsCompare ModelsPricingDevelopers · APIContext Window · Glossary

Key facts · as of 2026-10-16

  • Qwen3.5 397B A17B by Alibaba Qwen. BenchGecko score 69.4, rank 41 of 319 scored models (normalized average of public benchmark scores).
  • List price $0.55 input · $3.50 output per 1M tokens (as of 2026-10-16).
  • Sold by 10 providers (as of 2026-10-16): Alibaba $0.39 in / $2.34 out · DeepInfra (fp8) $0.45 in / $3.00 out · Parasail (fp8) $0.50 in / $3.60 out · AtlasCloud (fp8) $0.55 in / $3.50 out · DigitalOcean $0.55 in / $3.50 out · and 5 more. Every provider

How to cite · data as of 2026-10-16

Qwen3.5 397B A17B · benchmarks, pricing and providers. BenchGecko, data as of 2026-10-16. https://benchgecko.ai/model/qwen3-5-397b-a17b

Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP