OpenAI Says GPT-6 Astra Now Uses Less ChatGPT Subscription Usage on Long Tasks
OpenAI executive Tibo says recent improvements to GPT-6 Astra reduce the subscription usage drawn by long-tail work for power users signed in with a ChatGPT account. The approved announcement claims up to 3–4x less usage and says there is no change in quality, but it does not publish implementation details, a new quota, or an API billing change.
OpenAI executive Tibo says recent improvements to GPT-6 Astra can reduce the subscription usage drawn by long-tail work for power users signed in with a ChatGPT account. In an approved-source post, he described the change as a usage improvement with no change in quality and said the long tail can see up to 3–4x less usage drawn from the subscription. (Tibo on X)
That is a meaningful change for people who use Astra for extended or unusually demanding work, but the announcement is deliberately narrow. It does not explain which internal improvements changed consumption, publish a new plan allowance, or announce a new API price.
Key takeaways
- OpenAI says GPT-6 Astra usage is more efficient on the long tail for power users signed in with a ChatGPT account.
- The approved announcement says the change can result in up to 3–4x less subscription usage on those tasks.
- Tibo says there is no change in quality.
- The post does not establish a new quota, reset cadence, plan entitlement, or API billing change.
- “Up to” is an upper-bound claim, not a promise that every Astra task will use one-quarter to one-third as much.
What the announcement actually says
The post says OpenAI made “some improvements that improve usage on the long tail for power users of Astra when logged in with your ChatGPT account.” It then says there is “no change in quality” and that the change can be a “pure win” that, on the long tail, results in up to 3–4x less usage being drawn from the subscription. (Tibo on X)
The wording leaves the implementation unspecified. It does not say whether the improvement comes from changes to context handling, caching, routing, generation, tool use, or another part of the product. Those details should not be inferred from the usage result alone.
This is usage efficiency, not a new allowance
OpenAI’s current help documentation describes ChatGPT Work and Codex usage as dependent on the model, task size, reasoning effort, and amount of work performed. It also says a longer task can use more of an allowance than a short task, so a fixed message count is not a reliable measure of remaining usage. (OpenAI Help Center)
Against that backdrop, the Astra announcement is best read as a change in consumption for some workloads. If the same long-tail task now draws fewer units from the included allowance, the user may be able to complete more work before reaching a limit. That does not mean the plan's published allowance has been multiplied or that every task receives the same reduction.
OpenAI’s help page also says that actual usage varies with the model, task, input and output size, reasoning settings, and Fast mode. Users should therefore compare their own account’s usage before and after the change rather than convert the “up to 3–4x” claim into a universal quota. (OpenAI Help Center)
What power users should do
If you use Astra for long-running work, keep an eye on the account’s usage indicator while repeating a representative workflow. A useful comparison is the same task shape, similar context, and the same reasoning setting before and after the update. Avoid treating a short prompt as a test of the long-tail behavior described in the post.
If the task still consumes the allowance faster than expected, OpenAI’s help documentation says that usage depends on the workload and the experience where it runs. Check the model picker, usage details, and any limit banner tied to your account rather than relying on a generic message count. (OpenAI Help Center)
What the post does not establish
| Question | What is established | What remains unknown |
|---|---|---|
| Quality | Tibo says there is no change in quality | No evaluation or measurement is published in the post |
| Usage | Up to 3–4x less subscription usage on the long tail | The affected task types, average reduction, and rollout details |
| Plans | The post refers to users logged in with a ChatGPT account | A plan-by-plan eligibility list or new quota |
| API | Nothing about API billing is announced | Whether any API behavior changed |
| Implementation | OpenAI says improvements were made | Which technical mechanisms produced the reduction |
Keeping those boundaries matters because usage efficiency, included capacity, and API pricing are different product surfaces. OpenAI’s durable documentation confirms the general task-dependent usage model, while the approved X post is the source for this specific Astra efficiency claim. (OpenAI Help Center; Tibo on X)
Bottom line
OpenAI says GPT-6 Astra now consumes less ChatGPT subscription usage on the long tail for power users signed in with a ChatGPT account, with up to 3–4x less usage claimed for those workloads and no quality change. Treat it as a workload-efficiency improvement, not as a new quota or API pricing announcement, and use your own account’s usage indicator to see how the change affects your tasks.
Sources
- Tibo on X, “We've made some improvements that improve usage on the long tail for power users of Astra,” September 6, 2026: https://x.com/thsottiaux/status/2096717905614524491
- OpenAI Help Center, “ChatGPT Work and Codex”: https://help.openai.com/en/articles/20001275-chatgpt-work-and-codex
Read next
Keep building the workspace playbookOpenAI Launches GPT-6 Astra for Computer Use and Complex Work
OpenAI is rolling out GPT-6 Astra, a model focused on computer use, browsing, software engineering, cybersecurity, science, and complex professional workflows. Access starts with a limited set of organizations before expanding to ChatGPT plans and developer platforms.
OpenAI Resets ChatGPT Work And Codex Limits For Paid Users
OpenAI executive Tibo announced usage resets for paid ChatGPT Work and Codex users on August 8 and August 29, 2026, then said the five-hour limit would return for Plus accounts across both products on August 26. The approved posts do not publish a numeric allowance, a new API billing rule, or details beyond that plan and product scope.
Codex Permission Profiles: Least-Privilege Controls for Local Agent Work
OpenAI shipped Codex permission profiles in beta -- reusable, inheritable policies that replace the coarse sandbox_mode/sandbox_workspace_write combo. A profile binds OS-enforced filesystem read/write/deny rules (down to **/*.env) to per-domain network and Unix-socket rules. Enterprise admins get fail-closed allowlists via requirements.toml. Profiles govern local sandboxed command execution only, not MCP servers, app connectors, browser, or cloud.
Frequently Asked Questions
What did OpenAI change in GPT-6 Astra?
OpenAI executive Tibo said the team made improvements to GPT-6 Astra that help usage on the long tail for power users who are logged in with a ChatGPT account. The post does not describe the implementation.
How much less subscription usage does the post claim?
The approved post says the improvement can result in up to 3–4x less usage being drawn from the subscription on the long tail. That is a claimed upper bound, not a promise that every task will use one-quarter to one-third as much.
Did GPT-6 Astra get a quality change?
No quality change was claimed in the announcement. Tibo described the improvement as a usage win and explicitly said there was no change in quality.
Is this a new ChatGPT quota or pricing change?
The post does not announce a new quota, a reset schedule, a plan change, or a price change. It describes how much subscription usage some long-tail Astra work may draw after the improvement.
Does this change OpenAI API pricing?
No API change is established by the approved announcement. It specifically discusses power users logged in with a ChatGPT account and subscription usage, not API token prices or API rate limits.