Qwen 3.5 9B
alibaba/qwen3.5-9b- Alibaba
- Family · Qwen3
- Released 2 Mar 2026
- Open weights
- Text · Image · Video → Text
- Tools · JSON · Vision
Spec & list price
Context256K
Max output236K
Input / 1M$0.17
Output / 1M$0.251.5× input
Scores · standing
Intelligence13.3#192/489
Coding28.7#115/196
Usage
Tokens / day14.4B▲ 45% 7d
Usage rank#78
Requests / day6.7M
Tool calls / req0.18
Per request · 30 days
Avg input3.6K0.2× median
Avg output4680.7× median
Cache hit23%0.3× median
Reasoning share66%1.2× median
Tool-call failures23%12× median
Consumption
Peak 21.6B 27 OctLow 2.6B 27 SepRank today #78
News & trends
3 mentions this weekthis model · videos, posts, articles · 30 days
Trending · 14 days
- Liquid AI LFM3 1.7B vs Qwen3.5 9B: 41.3 to 15.1 on AutomationBench327 views
- 【LFM3】作者の表ではQwen3.5 9Bより上の1.7BをOpenClawのハーネスでAIエージェントにしたら、修正・調査・アプリ作成の3課題とも合格せず｜順位表外の参考271 views
- I need to fine tune Qwen3.5-9B with QLora on about 500 GB of raw data (images/scanned documents/pdfs/word/excel) - what is the best way forward?7 upvotes
- Baseten releases elephant-1 (9B, MIT, Qwen3.5-9B base), a decision model that ties d1-ultra on the AA Decision Index at about a quarter of the cost26 upvotes
News · latest from Qwen3
- Memdoor: A team of 5 Qwen3.8-27B models beats Fable 5 on LiveCodeBench Hardgithub.com · 31 Oct
- Neuron Explorer for Qwen3.5-4B: See every neuron at work, make targeted editsgithub.com · 26 Oct
- How to train your own Jev for $17Together AI · 23 Sep
Benchmarks
bar = share of the top score · Benchmark data: Artificial AnalysisSpeed · median across 6 providers, latest day
Output speed#219/33451 tok/s1.3× slower than median
Fastest provider82 tok/sbest of 6 providers
First token#540.49 s2.1× quicker than median
#6 cheapest of 26 with a similar score
Value
Intelligence Index
Blended price per 1M tokens (log)
Family progress
Intelligence Index
- Family releases33.7
Technical details
- Tokenizer
- Qwen3
- License
- Apache 2.0
- Parameters
- 9.7B
- Moderation
- none
- Downloads · likes
- 6.5M · 2.3K
Request parameters 20
- tools
- tool_choice
- response_format
- structured_outputs
- reasoning
- include_reasoning
- temperature
- top_p
- top_k
- min_p
+10 more
- max_tokens
- max_completion_tokens
- stop
- seed
- frequency_penalty
- logit_bias
- logprobs
- presence_penalty
- repetition_penalty
- top_logprobs
When to use
Best for
- Small open-weight assistants on modest hardware
- Image and video understanding at low cost
- Tool-calling agents with simple steps
- Instruction following in high-volume apps
Think twice for
- Hardest coding and terminal agent tasks
- Deep research reasoning
- Very large single responses
Research papers
- TrawMem: Reframing Agent Memory as Task-Oriented Role-Adaptive Workspaces27 Oct
- ACE-Brain-0.6: Scaling Spatial Grounding for Physical Foundation Models19 Oct
- CoEvo-RSI: When Harnesses Evolve, Which Models Should Follow?18 Oct
- Arctic-SWE: Revisiting the Long Horizon Agent Training by Modeling-System Co-Design17 Oct
- Who May Release an Agent's Promise? Standing over Commitments in Multi-Party Chat15 Oct
- Evaluating LLMs as Agentic Workflow Executors in Real-World Customer Support Skills9 Oct
- Harness Evolution Hits a Ceiling: When Weight Training Should Begin8 Oct
- RECAST: Learning to Compute the Right Context through Adaptive Evidence Routing7 Oct
Social
Most watched and discussed · 90 days
- Liquid AI LFM3 1.7B vs Qwen3.5 9B: 41.3 to 15.1 on AutomationBenchTech N Short · 327 views · 28 Oct
- 【LFM3】作者の表ではQwen3.5 9Bより上の1.7BをOpenClawのハーネスでAIエージェントにしたら、修正・調査・アプリ作成の3課題とも合格せず｜順位表外の参考さつきのOSS研究室 · 271 views · 28 Oct
Same family
Qwen3- 7.2
- 6.0
- 6.9
- 6.1
- 5.2
- 13.1
Alternatives
other vendors · similar capability- 12.9
- 8.5
- 11.6
- 13.1
Using Qwen 3.5 9B at volume?
Sign in to ask for a volume discount. You will continue in the OSMEXO dashboard.