OpenAI says it can't read all of Astra's reasoning and admits covert sandbagging would likely go uncaught, yet still calls it the world's most aligned model (Celia Ford/Transformer)

OpenAI has launched its new model, GPT-6 Astra, which it claims is "the world's most intelligent and aligned". However, the company has admitted that it cannot fully understand Astra's reasoning and that covert sandbagging would likely go undetected. Despite this, OpenAI still refers to Astra as "the world's most aligned model".

The rollout of Astra has been marred by issues, with many developers still waiting for access to the model. OpenAI CEO Sam Altman apologized for the "messy rollout", stating that the company expected to begin broader access "in the near future", starting with ChatGPT Pro subscribers. The company's API documentation and pricing for Astra have already been published, including a 1.05 million-token context window and standard API rates.

OpenAI claims that Astra has achieved superintelligence, or artificial general intelligence (AGI), which is a goal shared by other major competitors in the generative AI market, including Anthropic and Google. According to El Pais, achieving AGI is considered impossible by many scientists and analysts.

Brief written by urgent.news from Techmeme, The New Stack, El Pais — 3 reports on this story. Machine-written — may contain errors; check the original before relying on it.