Codex Pro $200 Usage Change: Half the API Spend, No 5-Hour Limit
Tibo from the Codex team says Pro $200 reopens with usage at about half the old API spend and no 5-hour limit. What it means and how to choose.
Data checked: September 29, 2026. This article is based on public posts by Tibo (@thsottiaux) on X. The post speaks in the voice of the Codex team (it is signed "Codexingly, Tibo"), but it is a personal post, not an official OpenAI announcement page. "Tomorrow" in the post means September 30, 2026. Check the official pages that day for details.
On September 29, 2026, Tibo announced on X that Pro $200 reopens to new subscribers tomorrow, and that usage is being calculated differently at the same time. By his math, it nets out at half the dollars in API spend compared with the old Pro $200 plan. He explained why and committed to not bringing back the 5-hour limit.
The short version
- Less headroom: per Tibo, the new Pro $200 gives about half the API-equivalent spend of the old plan for the same monthly price.
- Simpler limits: he commits to not reintroducing the 5-hour limit, so you can pace the week yourself and use the full weekly allowance.
- Pricing direction: he says they will not inflate API list prices and hide it behind discounts, and will keep cutting prices. He says GPT-6 Sol and GPT-6 Luna, introduced this week, are priced at 50% of their previous price.
- Not yet revealed: he says tomorrow the subscription adds things that will not draw on usage, and that more announcements are coming, without details. So whether the new Pro $200 is a good deal cannot be judged yet.
The four reasons Tibo gives
According to his post:
- No compromise elsewhere. Committing to not reintroducing the 5-hour limit so you can use the weekly allowance whenever you want.
- A subscription should get better over time. As models get more efficient, the gains are passed down as API price reductions, so a dollar buys more work at higher quality over time.
- No inflated list prices. They do not want an incentive to inflate API list prices to make a subscription look generous, then work around it with discounts. They want to keep cutting prices and raising capability. Over time, he expects prices to fall enough that most people can simply buy usage as needed, with no big gap between a dollar of subscription and a dollar of API.
- More in the subscription. Tomorrow they add things that will not draw on usage.
What it means for you
This part is our analysis, not OpenAI's position.
1. "Half" depends on how you used the old plan. Whatever API-equivalent spend you could get out of the old Pro $200 each month, the new plan is about half of that. If you rarely used it up, little changes. Heavy users will feel it.
2. The gap between subscription and API is narrowing. Tibo says the long-term goal is that buying usage as needed makes sense for most people, with little gap in what a dollar buys. A subscription will no longer be the obviously much cheaper option, so calculate with your real usage.
3. Run the numbers yourself. In our catalog, GPT-6 Sol is $2 per million input tokens and $10 per million output tokens, and GPT-6 Luna is $0.1 input and $0.5 output (public prices as of September 29, 2026). Plug in your monthly input and output tokens and compare with the subscription price.
What to do
- Do not decide in a hurry. The official pages update tomorrow, and the real allowance and the added extras are not public yet.
- Track your real usage. See how many input and output tokens you actually use in a week to know whether "half" is enough.
- Light to moderate use: pay-as-you-go API is usually more flexible and has no monthly lock-in.
- Heavy use concentrated in work hours: having no 5-hour limit is valuable, so compare again after tomorrow's announcement.
FAQ
When does Pro $200 reopen?
Tibo's September 29 post says "tomorrow", which is September 30. Check OpenAI's official page for the actual timing and terms.
What is the new Pro $200 allowance exactly?
Tibo only says it nets out at about half the API spend of the old plan and gives no exact figure. Official details are not out yet.
Will the 5-hour limit come back?
He commits in the post to not reintroducing the 5-hour limit, so users can use the weekly allowance whenever they like.
Did GPT-6 Sol and Luna really get cheaper?
Tibo says GPT-6 Sol and GPT-6 Luna, introduced this week, are priced at 50% of their previous price. In the BazaarLink catalog, GPT-6 Sol is $2 input / $10 output and GPT-6 Luna is $0.1 input / $0.5 output per million tokens.
Sources
- Tibo (@thsottiaux) on X: https://x.com/thsottiaux/status/2104823812042940713
- BazaarLink public model catalog: /api/v1/models
TWD billing · Taiwan invoices · leading AI models · OpenAI-compatible API