IMPLICATOR
OpenAI is giving paid ChatGPT subscribers one banked usage reset for every day they remain without GPT-6 Astra. The resets accrue because the model is included in existing paid-plan allowances but did not arrive for many subscribers when the rollout began. The offer addresses a gap between the access OpenAI announced and what its customers could actually use.
What Changed
- OpenAI is giving paid ChatGPT users one banked usage reset for every day they go without GPT-6 Astra, counting from September 3, 2026. The measure was set out at 11:12 p.m. that day, with the first reset due roughly three hours later.
- Astra was announced on September 3 as reaching all ChatGPT Plus, Pro, Business and Enterprise users "over the coming days." Many paying subscribers did not get it, and Sam Altman apologized early on September 4 for what he called a "messy" rollout.
- By September 4, Altman said Astra was available to all Pro, Enterprise and Business Premium users. Plus was absent from that list, and OpenAI had not confirmed Plus access as of September 5.
- For API developers at the September 3 launch, Astra cost $10 per million input tokens and $50 per million output tokens, against $5 per million input tokens for GPT-5.6 Sol Standard.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
The rollout sequence
OpenAI announced Astra on September 3, saying the model was rolling out that day to a limited set of organizations. The first group included customers in Daybreak, its application-based cybersecurity program. The company said Astra would reach all ChatGPT Plus, Pro, Business and Enterprise users "over the coming days," alongside its API, Microsoft Azure and AWS Bedrock.
That wording set out the intended destination without promising immediate access to each account. Sam Altman posted the launch at 7:49 p.m. on September 3. Minutes after announcing the model, he disclosed "a little snag" with deployment of the announcement blog post itself.
Subscribers continued asking when they would receive a model advertised as part of their plans. Early on September 4, Altman apologized for what he called a "messy" rollout.
FREE AI BRIEFING · WEEKDAYS
Track what OpenAI actually ships, and when.
Get the AI stories shaping the day, with concise context from San Francisco. The briefing takes about five minutes and arrives at 4:45 a.m. Pacific, 7:45 a.m. Eastern.
Free. No hype. Unsubscribe anytime.
Resets for the wait
At 11:12 p.m. on September 3, Thibault Sottiaux, an OpenAI member of technical staff who leads Codex, set out the compensation. A paid ChatGPT user would receive one banked reset for each day without Astra, starting September 3. The first reset was due roughly three hours after his post.
Sottiaux said the team was "moving mountains" to expand access. Astra carries no added charge within the subscriptions that include it, and users can purchase credits for more usage. Pro, Business and Enterprise plans are also due to receive an Astra Pro tier. "When we screw up, we try to make it right," Altman said in his apology.
For developers using the API at the September 3 launch, Astra cost $10 per million input tokens and $50 per million output tokens, against $5 per million input tokens for GPT-5.6 Sol Standard.
Why OpenAI held access back
A company spokesperson said the initial release gave OpenAI time to add compute capacity because Astra is "a very large model." Early findings that Astra might reach the Critical cybersecurity threshold prompted OpenAI to delay parts of development for several weeks while strengthening protections. On September 1, the company said the model had met the threshold, making Astra the first OpenAI model designated at that level under its Preparedness Framework. The public release refuses advanced offensive cyber tasks, while selected defenders receive broader access through the Daybreak program.
Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.
Sanchit Vir Gogia, chief analyst at Greyhound Research, challenged the idea that the designation marked a sudden capability jump. "Astra's capability did not change between 10 August, when OpenAI said Critical capability could not be ruled out, and September 1, when it said the threshold was met. The testing changed. The model did not."
Plus subscribers were still waiting
By September 4, Altman said Astra was available to all Pro, Enterprise and Business Premium users. His list did not include Plus. Enterprise access was off by default at launch and still required a workspace administrator to enable it.
OpenAI had not confirmed that Plus subscribers had Astra as of September 5. Altman also declined to promise a delivery date. Asked early on September 4 what "near future" meant, he replied: "i am hopeful that you can use it this weekend! but can't promise yet."
Frequently Asked Questions
What compensation is OpenAI giving subscribers who cannot use GPT-6 Astra?
One banked usage reset for every day a user lacks access to Astra on a paid ChatGPT plan, counting from September 3, 2026. OpenAI set out the measure at 11:12 p.m. that day, and the first reset was due roughly three hours later.
Which ChatGPT plans had GPT-6 Astra?
By September 4, Sam Altman said Astra was available to all Pro, Enterprise and Business Premium users. Plus did not appear in that list, and OpenAI had not confirmed Plus access as of September 5. Enterprise access was off by default and required a workspace administrator to enable it.
Does GPT-6 Astra cost extra on a ChatGPT subscription?
No. Astra is included within existing paid-plan allowances, and users can purchase credits for additional usage. Pro, Business and Enterprise plans are also due to receive an Astra Pro tier.
What does GPT-6 Astra cost through the API?
At the September 3 launch, $10 per million input tokens and $50 per million output tokens, against $5 per million input tokens for GPT-5.6 Sol Standard.
Why did OpenAI stagger the release?
Early findings that Astra might reach the Critical cybersecurity threshold prompted OpenAI to delay parts of development for several weeks while strengthening protections. On September 1 the company said the model had met the threshold. A spokesperson also said the limited start gave OpenAI time to add compute capacity.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai
