AI
Altman acknowledges 'messy' Astra rollout as paying users face delayed access to GPT-6
OpenAI's new flagship model launched Thursday with access restricted to enterprise customers, frustrating paid subscribers who were left waiting for their turn.
The Verge
OpenAI's long-awaited GPT-6 Astra arrived Thursday with a rollout plan that left many paying customers frustrated. CEO Sam Altman soon took to social media to apologize for what he called a "messy" launch, acknowledging the disappointment among subscribers locked out of immediate access.
The staggered rollout fuels frustration#
The company initially made Astra available only to enterprise clients with access to its Daybreak cybersecurity platform, with plans to expand to Plus, Pro, Business, and Enterprise tier subscribers over the following days. The decision to prioritize business customers over individual paid users—especially Pro plan holders accustomed to first access—sparked immediate backlash on social media.
Subscribers complained publicly on X (formerly Twitter) about being excluded from a product they'd long anticipated. The frustration was particularly acute among those paying premium prices for early access to new features.
Unclear timeline and compensation offer#
Neither Altman nor other OpenAI representatives initially provided a specific date for when broader access would arrive. Altman suggested users "might" gain access over the weekend but stopped short of guaranteeing it. The company did offer a partial peace offering: Thibault Sottiaux, a Codex engineering lead, announced that paid customers would receive one free token reset for each day they lacked access to Astra.
Additional launch complications#
The rollout faced other hiccups beyond access delays. Altman acknowledged a technical issue deploying the company's official announcement post, among other problems during the launch day.
The new model has also drawn scrutiny from AI safety advocates. OpenAI revealed that Astra's reasoning processes are harder to interpret than other models in its lineup—a development that raised concerns given the company's recent security incidents and its stated belief that interpretability is critical for safe AI development.
A pattern of tumultuous releases#
This wasn't OpenAI's first rough product launch. The company's rollout of GPT-5 last year was similarly fraught with technical failures and user backlash, particularly after the sudden removal of the popular GPT-4o model. Altman himself acknowledged at the time that OpenAI "totally screwed up some things" during that release.

