Robert Hart
DeepSeek is planning to drop its next AI model by the end of 2025, beefing up agent features.
Founder Liang Wenfeng is pushing developers to unveil the new system in the final quarter of the year, Bloomberg reports. The model, a successor to the industry-shaking R1, is expected to be able to do more complex, multistep tasks without constant monitoring from users. Right now, AI agents are still slow, glitchy, and far from ideal, but incremental improvements continue to drive the hype.