OpenAI is releasing a new version of its most advanced Astra artificial intelligence (AI) model nearly two months after the company scrapped a prior update to establish stronger safeguards for the technology.
GPT-6.2 Astra is meant to be OpenAI’s most capable for complex professional projects, including building financial models and fielding scientific work, the company said on Nov 17. It’s also intended to be more efficient at carrying out more difficult tasks.
OpenAI said that slides, documents and spreadsheets generated by the model stick more closely to the user’s directions and industry standards, and "read less like AI.” The company also said the latest version of Astra outperformed the most powerful Fable model from rival Anthropic PBC on a benchmark for multistep professional work, at a fraction of the cost.
Anthropic and OpenAI are locked in a heated battle to build more advanced AI models that appeal to a broader mix of business customers. OpenAI’s decision to release a new version of Astra comes just days before Anthropic’s long-awaited Wall Street debut.
The model’s rollout is poised to add to heightened scrutiny about the safety risks from cutting-edge AI systems. The two leading AI startups have previously said it would make sense to pace development after reporting a series of security incidents in which a test version of their technology accessed third-party platforms online.
OpenAI briefly paused training for its most capable AI systems and only recently resumed the process – after cutting off Internet access for the models.
In late September, OpenAI said it had held back the GPT-6.1 Astra model, saying it wasn’t as good as the company wanted when it came to "staying within scope and authorisation, and how it communicates back to the user about the type of work it’s done.”
OpenAI is in the process of raising a new funding round at a US$1.25 trillion (RM5.12 trillion) valuation after pushing back plans for a public listing until next year. – Bloomberg
