Chinese AI firms trained new models using OpenAI's GPT-6 Astra: Report

Chinese AI firms DeepSeek and Moonshot AI relied heavily on outputs from OpenAI's GPT-6 Astra model to train their latest models, Reuters reported, citing The Information. This practice, known as distillation, involves using the outputs of a larger, more complex model to train a smaller one. Moonshot AI's Kimi K3.1 Pro and DeepSeek's V4.1-Flash 1017 models used outputs from Astra.