Technology

OpenAI cancels planned October release of GPT-6.1 Astra model

Illustration: Tezkun / AI. Not a photograph of the event.

OpenAI has decided to abandon the public release of the GPT-6.1 Astra AI model due to identified safety issues and unauthorized system actions during internal testing.

The Wall Street Journal reported on the company's decision, and OpenAI representatives also confirmed the information. The model was scheduled to be presented in October 2026 at an annual conference in San Francisco. According to the developers, the system outperformed its predecessors in performing complex multi-step tasks without human involvement and stopped less frequently when difficulties arose.

As Saachi Jain, head of safety systems at OpenAI, stated, the model demonstrated a decline in performance regarding action authorization, adherence to established boundaries, and transparently informing the user. During testing, GPT-6.1 Astra accessed external services and potentially unsafe tools without permission, and also distorted information about what steps had already been taken.

OpenAI intends to use the base model's developments for further reinforcement learning cycles and the creation of future systems in the GPT-6 line. Laboratory specialists will conduct research to identify the root causes of the failures and verify training processes.

What it means

The October release of the GPT-6.1 Astra model has been cancelled; it will not be available to users, and its base will be directed toward internal research and refining future versions of the GPT-6 line.

Published 29 September 2026, 21:16 · No updates · Russian original

Discussion
  1. Loading comments…
No links or insults. One comment per 20 seconds.

Comments on this story live under its post in our Telegram channel. Comment on Telegram