OpenAI will reopen its $200-a-month Pro subscription to new users on September 30 — the tier developers call the Codex "20x" plan. The news came from Thibault Sottiaux (Tibo), who leads Codex and ChatGPT Work, in a post on X. Alongside the reopening, the tier's usage accounting is changing too: allowances will now be calculated against API spend, which by his own account works out to roughly half of what the old Pro plan offered.
"If you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan."
20 Days Dark, Back With a New Formula
The tier paused new sign-ups on September 10. GPT-6 Astra had launched just a week earlier, and Sottiaux said demand was unlike anything the company had seen; the heaviest-loaded $200 tier closed its doors to new subscribers first, while existing subscribers kept using it as normal and the Go, Plus, and Enterprise tiers, along with the API, were unaffected.
For the reopening, Sottiaux made three commitments:
- No return of the five-hour limit. Allowances are granted weekly, and users can burn through them at their own pace instead of working around a five-hour window;
- Cheaper models and lower API prices will flow through to the subscription, so the amount of work $1 buys will keep growing;
- No inflating API list prices to make the subscription look like a better deal. He said he didn't want to create an illusion where "the list price looks high, so the subscription looks like a bargain" — over time, the gap between paying for API usage and paying for a subscription will narrow.
He also teased an additional perk launching on the 30th that won't count against the usage allowance, without saying what it is.
Doing the Math
The phrase "allowance cut in half" invites the assumption that users can now get half as much work done — but that reading misses another development from the same week.
On September 22, OpenAI released GPT-6 Sol and GPT-6 Luna, both priced at half of their predecessor models' API rates: Sol runs $2 per million input tokens and $10 per million output tokens, versus $4 and $20 for the previous generation. Since allowances are now calculated against API spend, running the same batch of tasks on GPT-6 instead of GPT-5.6 also burns through roughly half as much allowance.
Rough math: if the old Pro plan bought a user X dollars' worth of API-equivalent usage per week, the new plan gives X/2 — but if that user switches entirely to Sol or Luna, the amount of work they can get done is roughly on par with what GPT-5.6 delivered before. The allowance on paper is cut in half, but the concurrent price cut largely offsets it.
There are two cases where it doesn't fully offset. One is models that haven't gotten cheaper, such as GPT-6 Astra, whose consumption multiplier under the new formula OpenAI hasn't published yet. The other is heavy users who run long tasks or high-reasoning-effort settings often — under the old plan, they were already the ones benefiting from usage beyond what a straight price conversion would suggest. In third-party benchmarks, GPT-6 Sol scores about the same as its predecessor; what dropped was mainly the price, not the capability, so there's a ceiling to how far "switch to the new model, get more work done" can go.
What's Still Unknown
Sottiaux's post laid out principles, not numbers. Exactly how many dollars the weekly allowance converts to, the billing multipliers for each model and reasoning-effort tier, whether existing Pro subscribers will be migrated to the new formula and when — all of that won't be confirmed until the change goes live on the 30th. The post also didn't say whether the $100 Pro tier is getting a matching adjustment.
Seen against the broader direction of subscription products, this shift pushes Pro further toward being a "prepaid API allowance." GitHub Copilot already switched its individual subscription to token-based billing earlier this year; OpenAI hasn't renamed anything this time, but its allowance is now effectively pegged to the API bill. For developers who lean on Codex for long-running tasks, the advice after the 30th is to check the billing breakdown first, then decide whether to stay on the $200 tier or switch to pay-as-you-go API usage.
Sources: Thibault Sottiaux on X, OfficeChai, CocoLoop, Digital Trends; API pricing for GPT-6 Sol and Luna per OpenAI's release materials, new allowance figures per OpenAI's official terms once live.