Qwen3.6-27B: A 27B dense model achieves flagship-level coding capability
Original title:Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model
AI summary
Qwen released the open-source model Qwen3.6-27B. This 27B-parameter dense model surpasses the previous-generation MoE flagship Qwen3.5-397B-A17B, with 397B total parameters and 17B active parameters, on coding benchmarks. Model size shrinks sharply from 807GB to 55.6GB; a 16.8GB quantized version can run locally, generating at about 25 tokens/s. Tests show it can generate complex SVG graphics code, demonstrating flagship-level coding capability.
Key Takeaways
- Qwen released the open-source model Qwen3.
- This 27B-parameter dense model surpasses the previous-generation MoE flagship Qwen3.
- 5-397B-A17B,with 397B total parameters and 17B active parameters,on coding benchmarks.
- Model size shrinks sharply from 807GB to 55.
- 8GB quantized version can run locally,generating at about 25 tokens/s.
Quick Facts
- Published
- Updated
- Source
- Simon Willison 博客
- Category
- Model Releases
- AI score
- 71
通义千问 3.6 27B 用 55GB 瘦身换来了超 397B MoE 的编码能力,Simon 亲手跑出的 SVG 效果惊艳,2026 年最值得本地部署的开源小模型。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Qwen3.6-27B: A 27B dense model achieves flagship-level coding capability. April 23, 2026. https://new.nat6.net/en/news/cmoczwjl40046slkqqx0d9t1v/