OpenAI’s new GPT‑6 Astra targets the enterprise with full‑stack AI agents

On September 27, OpenAI introduced GPT‑6 Astra, a new large language model aimed squarely at enterprise customers. Built on the existing GPT‑6 family, Astra is tuned for building production‑grade applications and AI agents that can automate complex business workflows end to end.

Astra ships with native “computer use” abilities: it can operate a virtual desktop in the cloud, log into internal business systems, and orchestrate multiple external APIs asynchronously. In practice, that means a single agent can read emails, gather data from an ERP system, update records, and generate reports without constant human intervention.

OpenAI is also rolling out a revamped Responses API around Astra. Developers can fine‑tune safety level, tone, and risk tolerance through granular steering controls, making the same model behave like a conservative assistant or a bold automation agent. While this flexibility is attractive to enterprises, it also raises the stakes: misconfigured agents could have broad impact, so logging, auditing, and safe rollback mechanisms will be critical for real‑world deployments.