According to @deepseek_ai, DeepSeek announced DeepSeek-V3.2-Exp as an experimental model built on V3.1-Terminus. Source: @deepseek_ai on X, Sep 29, 2025, https://twitter.com/deepseek_ai/status/1972604768309871061. It debuts DeepSeek Sparse Attention (DSA) to deliver faster, more efficient training and inference on long context. Source: @deepseek_ai on X, Sep 29, 2025, https://twitter.com/deepseek_ai/status/1972604768309871061. The model is now live across App, Web, and API channels. Source: @deepseek_ai on X, Sep 29, 2025, https://twitter.com/deepseek_ai/status/1972604768309871061. API prices have been cut by more than 50% as part of the release; key quantifiable catalysts for traders are the model’s availability and the 50%+ price reduction disclosed in the announcement. Source: @deepseek_ai on X, Sep 29, 2025, https://twitter.com/deepseek_ai/status/1972604768309871061.