OpenAI's GPT-6 Astra hits rough launch as developers await API access

OpenAI unveiled GPT-6 Astra on Thursday, unveiling an advanced language model that supports a 1.05 million-token context window and up to 128,000 output tokens. However, the debut stumbled: many developers and users eager to test the system immediately found themselves unable to access it.

CEO Sam Altman acknowledged the misstep publicly, writing on X that he was "sorry for the messy rollout" and noting that neither API customers nor ChatGPT subscribers had yet gained broad access. He pledged that the wider release would begin in the near future, with ChatGPT Pro subscribers receiving priority.

What happened during launch#

The situation left developers in an awkward position. OpenAI had already published Astra's full API documentation and pricing—set at $10 per million input tokens and $50 per million output tokens—but the endpoint itself remained unavailable. Reviewers could see the model's impressive benchmark results but couldn't validate those claims against real-world use cases.

Thibault Sottiaux, OpenAI's engineering lead for Codex, provided some context on the engineering challenges. He explained that the rollout would "take a few days" and that OpenAI is deploying new infrastructure systems while scaling up computational resources to meet demand. "Many novel systems will operate at scale for the first time" during this expansion, he noted, indicating that the company is essentially bringing massive new capacity online simultaneously.

Compensating waiting customers#

To offset the delay, OpenAI introduced what it calls a "banked reset" for paid ChatGPT subscribers—one additional reset credit for every day they remain unable to access Astra, beginning September 3. These resets, which have been used previously with other OpenAI products, allow users to refresh their usage allocations without permanently raising their limits.

Notably, OpenAI has not announced a similar compensation scheme for developers waiting for API access, leaving that segment of users without any official acknowledgment of the rollout impact.

Lingering questions#

The delayed rollout raises unresolved questions for the developer community. OpenAI has committed to rolling out access to ChatGPT Plus, Pro, Business, and Enterprise tiers, along with the API, Microsoft Azure, and AWS Bedrock "over the coming days," but has provided no specific timeline for when general API availability will arrive or whether it will come with usage restrictions.

Additionally, reporting from The New Stack has surfaced that Astra's API includes new safety-triggered interruptions that can mimic timeout errors, a change that could significantly affect developers building production systems on top of the model. OpenAI has not yet responded to detailed questions about the rollout timeline or these technical concerns.