OpenAI introduced GPT‑6 Astra last week, positioning it as the most capable model for professional tasks. The model is now live in ChatGPT Work, Codex, and via the API.
What Astra promises
Astra is built to handle computer use, browsing, software engineering, cybersecurity, and scientific work without the need for custom integrations. OpenAI says it can operate inside existing applications even when those apps lack an API, letting teams deploy AI from day one.
Early customer signals
Within days of rollout, OpenAI reports customers using Astra to optimize GPU workloads, spot discrepancies in financial statements, and generate on‑brand slide decks. Internally, the company used Astra to turn three hours of multicamera footage into a demo video that earned 550 k views in four days. The same internal test found a 25× reduction in turn latency after swapping memory allocators, while peak memory use rose by roughly 30 %.
Cost efficiency
Astra is priced at $10 per million input tokens and $50 per million output tokens. OpenAI notes that the model is trained to complete tasks in fewer tokens with fewer retries, which should lower the cost per task compared with earlier models.
Safety upgrades
Business use cases demand tighter controls. Astra's internal safety benchmark showed unintended outcomes 89 % less often than GPT‑5.6 and 74.7 % less often than Claude 5.1. New admin tools let enterprises restrict website access, manage file uploads, and require approval before consequential actions. The model also meets OpenAI's Critical cybersecurity capability threshold, adding automated checks against misuse and unauthorized actions.
Zero Data Retention is available for eligible API customers on supported endpoints, subject to approval.
What this means for the market
The headline is clear: OpenAI is trying to make AI a drop‑in component for existing business software. If the speed and cost numbers hold up in real‑world deployments, Astra could squeeze out smaller competitors that still need heavy integration work. At the same time, the safety controls may ease corporate concerns about granting AI access to critical systems.
I'm skeptical about the "no‑prep" promise. Most enterprises have legacy data policies, compliance checks, and change‑management processes that rarely disappear overnight. Astra's performance numbers look impressive, but they come from OpenAI's own tests. Independent benchmarks will be the real proof point.
Nobody knows how quickly companies will adopt a model that can both write code and click through GUIs. The next few months should reveal whether Astra lives up to the hype or becomes another incremental upgrade.
Sources
- OpenAI, GPT‑6 Astra: The next generation in intelligence for work (Sep 9 2026) - https://openai.com/index/gpt-6-astra-next-generation-work