The Qwen3.5 Series 35B-A3B is a native vision-language model designed with a hybrid architecture that integrates linear attention mechanisms and a sparse mixture-of-experts model, achieving higher inference efficiency. Its overall...
Specifications
- Provider
- Qwen (Alibaba)
- Type
- Open-source / open-weight
- Modality
- text+image+video->text
- Context window
- 262,144
- Released
- February 25, 2026
Capabilities
Input: textInput: imageInput: videoOutput: text
Strengths
frequency_penaltyinclude_reasoninglogit_biaslogprobsmax_tokensmin_ppresence_penaltyreasoningrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_logprobstop_p
