OpenAI says new Pro $200 subscriptions are available at $200 a month with a lower included allowance than the previous version. TechCrunch reported that OpenAI paused new sign-ups and upgrades on September 10; the Help Center notice at the time said existing Pro $200 subscribers were unaffected. The restart timeline is less clear. On September 29, Thibault Sottiaux, who leads product work across ChatGPT and Codex, said reopening was planned for “tomorrow,” September 30. A later update to OpenAI’s Pro Help Center page that day said new subscriptions were available. Those sources do not establish when access actually returned, so the precise pause duration cannot be confirmed. Eligible existing subscribers keep their former allowance through October 29, 2026, then move to the lower allowance at the same price. Sottiaux separately described the new allowance as worth about half the old plan in API-dollar terms. That is a value estimate at API prices, not a count of tasks users can complete.

Sottiaux posted the API-dollar comparison on September 29, the day of OpenAI's DevDay in San Francisco. The event schedule listed a 10 a.m. Pacific keynote (17:00 UTC, 19:00 in Belgrade). A pre-event snapshot at 16:04 UTC showed the official livestream page saying the stream would begin at 10 a.m. Pacific. OpenAI's updated Pro Help Center page now says new subscriptions are available. It does not publish the API-dollar calculation or task counts.

The big change

  • What changed: New customers can subscribe to Pro $200 again, but OpenAI says their included usage is lower than under the prior version. Eligible existing subscribers keep the former allowance through October 29, then transition to the lower allowance without a price change.
  • Why it matters: Sottiaux says the new allowance represents about half the former plan's API-dollar value. That is a price-based comparison, not evidence of half as many tasks; work completed varies with model, workload and usage limits.
  • What to watch: OpenAI's Help Center confirms availability, price and the transition date, but does not publish the usage quantities or Sottiaux's API-dollar calculation. The September 29 post and later Help Center update differ on reopening timing, so the exact restart time remains unknown. Those figures and the account's actual limits are needed to assess the practical effect.

OpenAI scheduled the September 29 DevDay keynote for 10 a.m. Pacific (17:00 UTC, 19:00 in Belgrade). The event listing linked the official OpenAI livestream page; at 16:04 UTC, the page said the stream would begin at 10 a.m. Pacific.

His explanation is that OpenAI wants to avoid tying subscription value to API list prices that it may reduce as models become more efficient. He says GPT-6 Sol and GPT-6 Luna arrived at half the prior models’ API prices, and argues that lower rates can buy more work over time. That is OpenAI’s rationale, not a measured guarantee for every customer or workflow.

The distinction matters for Codex users. A lower API-equivalent allowance does not by itself say how many tasks, tokens or hours of work a subscriber receives. The result depends on model choice, input and output mix, caching, reasoning, tools and task length. OpenAI’s current Codex plan guidance says usage varies with those factors and directs customers to the usage dashboard for their account’s limits and reset times.

BIG CHANGE previously examined a separate claim that GPT-6 Sol cut Codex quotas in half. The analysis of the API-price math found that halving rates can halve API-equivalent estimates for identical token counts, without proving how a subscription meter changes. Sottiaux’s statement about the new plan is more specific: he says its allowance represents about half the old plan's API-dollar value. The actual new and former usage amounts and account-level meter mechanics remain unpublished.

What customers still need to know

TechCrunch reported that OpenAI paused new subscriptions and upgrades on September 10; OpenAI’s Help Center notice at the time said existing $200 subscriptions were unaffected. Sottiaux’s September 29 post said reopening was planned for the next day, September 30. But a later update to the same Help Center page on September 29 said new subscriptions were already available. Since the post described a planned date and the Help Center update did not state when availability changed, these sources cannot establish the actual restart time or the pause’s exact duration. The current Help Center says subscribers active at the eligibility cutoff, or during the preceding seven days, keep the old allowance through October 29, 2026, as long as Pro $200 remains active. After that, they move to the lower allowance while the price stays $200 a month. The page does not specify the cutoff date.

Sottiaux says the five-hour cap will not return on the $200 plan, allowing subscribers to use the weekly allowance when they choose. The post does not specify the weekly limit, whether the same usage pool applies across ChatGPT and Codex, or how users will see model-by-model consumption. It also withholds the names and terms of the additional subscription features.

For teams comparing the subscription with API billing, the API-dollar equivalent is only one input. The Help Center does not give numeric allowance amounts, detailed reset mechanics or a task-based comparison. Sottiaux's statement is the only source located for the precise “half” API-dollar comparison; the updated Help Center independently confirms a lower allowance, but not the ratio or calculation. OpenAI says more efficient models and lower API prices can increase work over time. That is the company's rationale, not a measured result for every user or workflow.

BIG CHANGE’s GPT-6.1 Astra safety hold report covers a separate September decision about delaying a model release; it does not establish these subscription terms.

Sources & further reading

  • Thibault Sottiaux's September 29 post — The only located source for the claim that the new allowance equals about half the former plan in API-dollar terms. X returned 403; the full text was retrieved through a contemporaneous repost and third-party mirror. The post says reopening was planned for September 30; a later Help Center update on September 29 says subscriptions were available, so actual timing is unresolved. OpenAI's Help Center confirms a lower allowance, not the ratio or calculation.
  • OpenAI DevDay 2026 event page — Confirms the event date, location, keynote schedule, free livestream, technical sessions and demos. It does not list Pro subscription terms.
  • OpenAI livestream page — At 16:04 UTC, 56 minutes before the 17:00 UTC start, it said the stream would begin at 10 a.m. Pacific, consistent with the schedule.
  • OpenAI Help Center: About ChatGPT Pro tiers — Updated later on September 29 to say new Pro $200 subscriptions are available at $200/month with a lower allowance. This conflicts on timing with Sottiaux's post that day saying reopening was planned for September 30. The current page does not give the actual restart time, numeric allowance amounts or confirm the API-dollar half comparison. Eligible existing users retain the former allowance through Oct. 29, 2026, then transition to the lower allowance at the same price.
  • TechCrunch: OpenAI puts Pro subscriptions on hold due to Astra demand — Contemporaneous September 10 report of the sign-up and upgrade pause, attributed to OpenAI product lead Thibault Sottiaux. The Help Center notice at the time said existing subscribers were unaffected; that notice has since been replaced by the current availability and transition terms linked above. The precise restart time and pause duration are not established by the available sources.
  • OpenAI Help Center: Using Codex with your ChatGPT plan — Describes the account dashboard and factors that affect Codex usage; it does not disclose the new Pro plan’s limits.
  • BIG CHANGE: Did OpenAI halve GPT-6 Sol’s Codex quota? — Earlier analysis of why API-dollar estimates and subscription quotas measure different things.
  • BIG CHANGE: OpenAI holds GPT-6.1 Astra after scope and authorization concerns — Separate same-day report on the delayed GPT-6.1 Astra release.