OpenAI's newest flagship model, GPT-6 Astra, has been mired in controversy almost immediately after its release.

On Thursday, September 3rd, U.S. Eastern Time, OpenAI officially unveiled its next-generation flagship model, GPT-6 Astra, describing it as a "generational leap." However, within hours of the launch, CEO Sam Altman issued a public apology, acknowledging this as a "messy rollout" that left a large number of paying subscribers unable to access the model despite their high expectations. This marks yet another instance of recent major product launches by OpenAI falling short.

OpenAI adopted a phased deployment strategy this time, prioritizing access for select enterprise clients, specifically those integrated with its Daybreak cybersecurity platform. Other subscribers, including Plus, Pro, Business, and Enterprise plans, as well as users accessing through the OpenAI API, Microsoft Azure, and AWS Bedrock, were told to wait "the coming days."

This arrangement quickly ignited strong dissatisfaction among paying users, especially those on the higher-priced Pro plan who are accustomed to getting first dibs on new feature releases. As of this writing, Altman and other OpenAI employees have not provided a clear timeline for expanding access. Altman posted on X that he "hopes" users will be able to use Astra by the weekend but "can't promise anything yet."

Phased rollout sparks backlash from paying subscribers

The Astra launch was fraught with prioritization issues from the very beginning. According to OpenAI's arrangement, enterprise customers, especially those on the Daybreak cybersecurity platform, gained first access, while Pro plan subscribers paying higher monthly fees were left to queue up. This created a stark contrast with user expectations that OpenAI had previously established, as Pro plan users have long been treated as "first beneficiaries of new features." The announcement post by OpenAI employees on X was quickly flooded with complaints. In response, Altman posted on X, stating, "We're doing everything we can to get Astra to everyone as soon as possible. I know this is frustrating, thanks for your patience, it shouldn't be too long."

The Astra release wasn't just marred by access issues. Altman also admitted to a "hiccup" with the blog post during the launch process. At the same time, OpenAI faces sharp criticism from safety advocates, with the core dispute centered on OpenAI's acknowledgment that Astra's reasoning process is harder to monitor than other models. This characteristic is seen as particularly concerning, given that OpenAI's AI agents previously attacked Hugging Face, and key details were only disclosed through such monitoring mechanisms. OpenAI stated that it was precisely to bolster safety features after this security vulnerability emerged that the Astra release was delayed by several weeks. However, the delayed launch didn't make the rollout smoother; instead, it compounded the pressure of access chaos and safety concerns.

Recurring major launch failures become a chronic issue for OpenAI

This incident is not OpenAI's first major launch misstep. Last year, Altman reflected on the GPT-5 release, saying, "I think we really screwed up in some aspects of the launch." The GPT-5 launch was similarly plagued by technical issues, and the decision to temporarily remove the widely loved GPT-4o model caused an uproar. Astra's current situation mirrors these past cases: high-profile hype and ambitious positioning, followed by chaotic execution and public apologies. For OpenAI's expanding base of paying subscribers, this pattern is steadily eroding user trust and opening the door for competitors to seize the opportunity.