DeepSeek-V3.2-Exp released with sparse attention for long-text efficiency and API price cut of over 50%
Original title:DeepSeek-V3.2-Exp 发布,训练推理提效,API 同步降价
AI summary
DeepSeek released the experimental model DeepSeek-V3.2-Exp, which introduces the DeepSeek Sparse Attention (DSA) mechanism on top of V3.1-Terminus, greatly improving long-text training and inference efficiency with almost no impact on output quality. Public benchmark performance is basically on par with V3.1-Terminus.
Key Takeaways
- DeepSeek released the experimental model DeepSeek-V3.
- 2-Exp,which introduces the DeepSeek Sparse Attention (DSA) mechanism on top of V3.
- 1-Terminus,greatly improving long-text training and inference efficiency with almost no impact on…
- Public benchmark performance is basically on par with V3.
- 1-Terminus.
Quick Facts
- Published
- Updated
- Source
- 公众号:DeepSeek(深度求索)
- Category
- Model Releases
- AI score
- 79
稀疏注意力在几乎不影响效果的前提下大幅降低长文本成本,API 降价 50% 让依赖长上下文的应用获得了更现实的部署条件。
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: DeepSeek-V3.2-Exp released with sparse attention for long-text efficiency and API price cut of over 50%. September 29, 2025. https://new.nat6.net/en/news/cmsdu3h2d015grofzq1cpc4oe/