DeepSeek and Moonshot AI reportedly trained recent models using GPT-6 Astra outputs

Chinese AI developers DeepSeek and Moonshot AI relied heavily on outputs from OpenAI’s GPT-6 Astra to train their latest models, The Information reported, citing people with direct knowledge, according to Reuters.

The Information identified Moonshot AI’s Kimi K3.1 Pro and DeepSeek’s upgraded V4.1-Flash 1017 as models that benefited from distilling Astra, Reuters wrote. Reuters said DeepSeek’s model launched Oct. 17. Distillation uses one model’s outputs to help train another. Reuters reported that AI companies argue the practice can let rivals reproduce advanced capabilities more quickly and cheaply. Its account did not specify how much Astra output either developer used or demonstrate what capabilities the models gained from it.