Qwen3.5-122B-A10B
All models
AlibabaAlibabaQwenReleased February 25, 2026

Qwen3.5-122B-A10B

262K context$0.260/M input$2.08/M output122B total, A10B active

The Qwen3.5-122B-A10B is a vision-language model built on a hybrid architecture combining linear attention mechanisms with a sparse mixture-of-experts model. It offers high inference efficiency and excels in both text and visual capabilities, outperforming earlier models like Qwen3-235B-2507 and Qwen3-VL-235B. It is designed for advanced reasoning and coding tasks, as well as long-context processing.

What is Qwen3.5-122B-A10B?

Qwen3.5-122B-A10B is an AI model from Alibaba that Agent Mag tracks for pricing, context window, modalities, benchmarks, and API compatibility. Builders can use this page to compare Qwen3.5-122B-A10B against other models for agent workflows and production deployments.

Model ID

The Qwen3.5-122B-A10B is a vision-language model built on a hybrid architecture combining linear attention mechanisms with a sparse mixture-of-experts model. It offers high inference efficiency and excels in both text and visual capabilities, outperforming earlier models like Qwen3-235B-2507 and Qwen3-VL-235B. It is designed for advanced reasoning and coding tasks, as well as long-context processing.

Architecture & Specifications
Architecture
Hybrid architecture with linear attention and sparse mixture-of-experts
Parameters
122B total, A10B active
Tokenizer
Qwen3
Released
February 25, 2026
Modalities
Input
textimagevideo
Output
text
Supported Parameters
frequency_penaltyinclude_reasoninglogit_biaslogprobsmax_tokensmin_ppresence_penaltyreasoningrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_logprobstop_p
Strengths
  • High inference efficiency
  • Advanced reasoning capabilities
  • Superior text and visual performance
  • Long-context processing
  • Hybrid architecture for optimized performance
Limitations
  • Lower performance compared to Qwen3.5-397B-A17B
  • Limited information on training data
  • High structured output error rates in some providers
  • Moderation responsibility left to developers
  • High tool call error rates in certain providers
Recommended Use Cases
Scientific reasoning and analysis
Coding and programming tasks
Vision-language applications
Long-context reasoning
Conversational AI in dual-control scenarios

Related content

Data enriched Apr 24, 2026. Pricing from OpenRouter API.