Model•active
DeepSeek-R1-0528
DeepSeek-R1-0528 is an open-weight reasoning model using the DeepSeek V3 mixture-of-experts architecture with 671B total parameters, about 37B activated parameters, and a 128K context window.
Official siteVerified Oct 4, 2026
Overview
The checkpoint is the selected representative for DeepSeek's R1 reasoning line. It targets reasoning-heavy math and coding work while remaining independently downloadable and self-hostable.
Model specifications
- Family
- DeepSeek R1
- Context
- 128,000 tokens
- Input
- Text
- Output
- Text
- Open weights
- Yes
- Capabilities
- Reasoning