DeepSeek · model release · downloadable model
DeepSeek-V3.2-Exp
DeepSeek releases a test model that is cheaper to run on long inputs, and cuts its prices.
In detail
DeepSeek released V3.2-Exp, an experimental model built on V3.1-Terminus that introduces DeepSeek Sparse Attention for long inputs. It cut API prices by more than 50% and published weights, a technical report and GPU kernels.
Sources
- Introducing DeepSeek-V3.2-Exp · DeepSeek · Primary source
Last checked 11 November 2026.