Tuesday, 31 March, 2026
DeepSeek Unveils New AI Model, Challenges OpenAI and Qwen
By Isha
Chinese AI firm DeepSeek has launched DeepSeek-V3.2-Exp, an experimental model that’s more efficient to train and better at handling long text compared to previous versions. It introduces a sparse attention mechanism, aiming to reduce compute costs. API pricing is also being cut over 50%, signaling stronger competition against Alibaba’s Qwen and leading global models.
Read full story at Hindustan Times