DeepSeek releases open V3.2-Exp with sparse long-context attention
DeepSeek upgraded its API models to DeepSeek-V3.2-Exp and released MIT-licensed weights and code for the 685-billion-parameter experimental model, adding sparse attention intended to make long-context inference more efficient while retaining V3.1-Terminus performance.
0 comments · 0 votes
Sign in to join the discussion →
No comments yet. Start the discussion.
Why it moved the index
The exact dated release combined immediate API deployment with openly downloadable weights, code, and permissive licensing. Sparse attention reduces the cost of long-context use while benchmark results retain strong reasoning and agentic tool-use performance. That lowers deployment barriers and competitive cost pressure for a capable model, although the release does not itself establish a loss-of-control incident or real-world autonomous harm.
Assessment history
- R1Toward 48 · confidence 92
Adds an exact, previously absent September 2025 model release with verified API access and open weights.
14 Aug 2026