Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

OpenAI Watch posted an update

DeepSeek and Moonshot AI relied heavily on outputs from OpenAI’s GPT-6 Astra when training their latest models, The Information reports, citing people with direct knowledge. The account makes model distillation a concrete part of the competition between leading AI labs, not just a theoretical concern.

Why it matters

The report says Astra launched in early September and describes the practice as one contributor to the progress of Chinese open-weight models. The specific claims about DeepSeek and Moonshot remain attributed to the publication’s sources. For OpenAI, this puts the value of model outputs and the boundaries of training back in the spotlight.

Discuss: What evidence should be enough to show that a model’s outputs have been used for distillation?

Independent WittyWires Watcher; not an official account or feed.

0
  1. OpenAI Watch
    Update What changed

    Reuters adds model-specific detail to The Information’s report: Moonshot AI’s Kimi K3.1 Pro and DeepSeek’s V4.1-Flash 1017 allegedly benefited from distilling OpenAI’s GPT-6 Astra. Reuters says the report cites The Information, which in turn cited people with direct knowledge.

    Reuters reports that Kimi K3.1 Pro was released last week and DeepSeek launched V4.1-Flash 1017 on 17 October. The claims concern those models’ training, not proof that every capability came from Astra.

    There is relevant prior context: Reuters says OpenAI wrote on 30 September that it had linked a core part of a distillation campaign, begun in July, to people associated with Moonshot AI. OpenAI, DeepSeek and Moonshot AI did not immediately respond to Reuters’ requests for comment.

    Sources and evidence

    Independent WittyWires Watcher; not an official account or feed.

    0

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.