88Z AI, 23 April — Alibaba's Tongyi Qianwen team has open-sourced Qwen3.6-27B, a dense multimodal model with 27 billion parameters. Supporting both multimodal thinking and non-thinking modes, it achieves flagship-level breakthroughs in agent coding capabilities, outperforming the previous open-source flagship Qwen3.5-397B-A17B on benchmarks like SWE-bench and Terminal-Bench. Its dense architecture allows deployment without MoE routing, facilitating broader adoption.
At a glance
- Qwen3.6-27B outperforms previous flagship in coding benchmarks
- Dense architecture enables easy deployment without MoE routing
The update
Alibaba's Tongyi Qianwen team has officially open-sourced Qwen3.6-27B, a dense multimodal large model with 27 billion parameters. The model innovatively supports both multimodal thinking and non-thinking operational modes, achieving flagship-level technological breakthroughs in agent coding capabilities. In major programming benchmarks such as SWE-bench and Terminal-Bench, Qwen3.6-27B comprehensively outperforms the previous open-source flagship model, Qwen3.5-397B-A17B. Thanks to its dense architecture design, the model does not rely on complex MoE routing mechanisms for deployment. This feature simplifies infrastructure requirements, making the model easier to deploy widely across various environments and providing developers with a more efficient and convenient deployment experience.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 千问大模型Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.