
'Sorry for the Messy Rollout': OpenAI Opens GPT-6 Astra to Most Paying Users After Confused Debut
A day after unveiling its new flagship, OpenAI widened access to ChatGPT Plus, Pro, Business and Enterprise tiers — apologizing for a staggered launch that left developers guessing about API availability.
OpenAI's biggest launch of the year got off to a stumbling start. A day after unveiling GPT-6 Astra, the company pushed the model to most paying users — ChatGPT Plus, Pro, Business, and Enterprise — while apologizing for a rollout that even insiders called messy.
What happened:
- The September 3 unveiling reached only a limited set of organizations at first, despite launch-day messaging that implied broad availability.
- Developer API access lagged the consumer rollout, leaving teams unable to test against the model many had cleared calendars for — prompting the "sorry for the messy rollout" mea culpa.
- Cyber-offensive capabilities remain gated behind access controls, a first for an OpenAI flagship after Astra hit the "Critical" cybersecurity threshold in the company's Preparedness Framework.
- Forbes noted the announcement itself had a "curious false start," with assets appearing and vanishing before the official reveal.
The stumble is cosmetic; the demand is not. Early benchmark claims — 99.9 percent on ARC-AGI-3, 97.6 percent on FrontierMath Tier 4 v2 — kept Astra the only story in AI for the week, and API traffic surged once access stabilized.
Still, the episode underlines a pattern: frontier launches have become logistically harder as safety gating, tiered access, and enterprise commitments multiply. The days of a model simply "going live" for everyone at once appear to be over — at the frontier, access itself is now a product decision.
Newsletter
Get Lanceum in your inbox
Weekly insights on AI and technology in Asia.


