DeepSeek-V3.2
Open Weights
DeepSeek
DeepSeek's December 2025 open-weight successor to V3.2-Exp, a reasoning-first model built for long-context tasks, agent workflows, and tool use. It introduces DeepSeek Sparse Attention (DSA) to cut long-context inference cost while keeping output quality on par with V3.1, with a 131K context window and 64K max output.
Strengths
- Reasoning-first design for long-context, agentic, tool-use tasks
- DeepSeek Sparse Attention (DSA) cuts long-context inference cost
- Open weights under permissive MIT license
- 131K context with 64K max output
- Very low per-token pricing
Caveats
- Text-only (no vision or audio input)
- DSA is a newer attention scheme with evolving tooling support
- Full-precision self-hosting demands substantial GPU memory
- Vendor
- DeepSeek
- Context Window
- 131,072 tokens
- Max Output
- 65,536 tokens
- License
- MIT
- Release Date
- 2025-12-01
- Modalities
- text
Capabilities
Vision
Audio
Video
Tool Use
Resources
Pricing
Open weights under the MIT License - free to self-host. DeepSeek no longer serves V3.2 via its API (the deepseek-chat / deepseek-reasoner endpoints were retired after 2026-07-24).