DeepSeek and Moonshot AI are said to have relied heavily on OpenAI’s GPT-6 Astra to train their latest models, according to a report by The Information. The account, cited by multiple outlets, points to Moonshot’s Kimi K3.1 Pro, released last week, and an updated version of DeepSeek’s V4.1-Flash, launched on 17 October.
The reporting frames the practice as a form of “distillation,” where outputs from a more capable model are used to train another system. One outlet describes this approach as a flashpoint amid intensifying US-China technology competition, with companies arguing it can help rivals reproduce advanced capabilities more quickly and cheaply.
Across coverage, the core claim remains consistent: both Chinese firms used GPT-6 Astra during training. Outlets differ mainly in how much background they provide about why the method matters, rather than on the underlying details cited about which models were involved.
