OpenAI expands access to GPT-6 Astra, targeting advanced AI agent tasks

OpenAI is expanding access to GPT-6 Astra following an initial restricted rollout of what the company describes as its most capable artificial intelligence model to date.

Built to handle complex, multi-step tasks, Astra is designed to perform strongly in areas including computer operations, software engineering, advanced mathematics and professional workflows.

The model has recorded high scores across a range of academic and technical benchmarks, including terminal-based evaluations and scientific reasoning tests, highlighting its ability to tackle tasks that require multiple stages of planning and execution.

Its initial release, however, was affected by temporary access bottlenecks for some paid subscribers as OpenAI worked to fine-tune safety controls and cybersecurity screening measures surrounding the new capabilities.

Astra is designed to move beyond conventional conversational AI by allowing models to interact with digital work environments and carry out tasks with greater independence.

The development could accelerate the shift toward AI agents capable of completing end-to-end workflows, from analysing information and writing code to operating software and executing complex digital tasks.

As access expands, attention is expected to focus on how reliably the model performs in real-world settings, as well as the safeguards required when increasingly capable AI systems are given greater control over computer-based operations.