AI News · 5 min read
ChatGPT Pro drops the Codex 5-hour cap: 10x, 20x, 50x Plus
OpenAI's Codex lead says ChatGPT Pro plans drop the 5-hour limit and give 10x, 20x or 50x Plus usage. What's official, what isn't, and how Claude compares.
What OpenAI said
On November 8, 2026 at 05:09 UTC, Thibault Sottiaux ("Tibo"), who leads Codex at OpenAI, posted a short clarification on X ↗:
"Saw some confusions, but all Pro plans have no 5-hour limits, and give the following usage compared to Plus. ✅ Pro 100 = 10X ✅ Pro 200 = 20X ✅ Pro 500 = 50X, access to ultrafast"
It drew more than 2.3 million views and 8,500 likes within a day. That reach makes sense. For anyone running long Codex sessions, the five-hour window has been the limit you actually hit.
This is a clarification, not a new price change. Sysnote ↗ points out that the allowance itself doubled on October 29. That announcement never published a full set of multipliers, which is where the confusion came from.
What's in OpenAI's own documents
Part of Tibo's post is backed by OpenAI's help center. Part of it, so far, rests on his word.
Confirmed in writing:
OpenAI's Work and Codex usage guide ↗ says: "Pro 100, Pro 200 and Pro 500 currently have no five-hour usage limit in Work and Codex. Your plan still has an included usage allowance."
The Pro tiers article ↗ lists the prices and says Ultrafast is only on Pro 500. It also says Pro 100 and Pro 500 "now include twice their previous Work and Codex allowance," and that a Pro 200 reduction announced on September 29 "is canceled."
Plus and Standard Business keep the five-hour window. For GPT-6 Astra on Plus, OpenAI estimates 5 to 45 local messages per five hours.
| Plan | Price | Usage vs Plus (per Tibo) |
|---|---|---|
| Plus | $20/mo | 1x, with a 5-hour window |
| Pro 100 | $100/mo | 10x, no 5-hour window |
| Pro 200 | $200/mo | 20x, no 5-hour window |
| Pro 500 | $500/mo | 50x, plus Ultrafast |
Not yet in writing: the 10x, 20x and 50x figures. El Mediático ↗ noted on November 8 that OpenAI's pricing page "doesn't show 10x, 20x or 50x in its text." OpenAI also doesn't say what 1x is in tokens or messages. And weekly limits still apply. "No five-hour limit" doesn't mean unlimited. OpenAI's Pro page says any unlimited access is "subject to abuse guardrails."
The fine print that changes your bill
Two details matter more than the headline multiplier:
Fast modes burn allowance faster. Per OpenAI, Fast mode uses included usage at 2.5x the standard rate. Ultrafast on GPT-6 Astra and GPT-6.1 Sol uses it at 8x, then charges credits at 6x once the allowance runs out. A Pro 500 subscriber who leaves Ultrafast on all day won't see anything like 50x Plus in practice.
Model choice still matters. Work and Codex share one allowance. OpenAI's Plus estimates show how far apart the models are: 5 to 45 messages per five hours on GPT-6 Astra, versus 15 to 160 on GPT-6.1 Sol. OpenAI's own guide adds that "Astra at Low effort can outperform Sol at High effort." It suggests trying Astra at Low or Medium if Sol at High has been working for you.
What people are saying
Most reactions came from Plus subscribers, not Pro ones. "Meanwhile, Plus users are still fighting the 5-hour restrictions 💀," wrote @jpshrodinger ↗. Eduardo Deodato Mendes asked Tibo to "keep the 5-hour limit, but please double Plus users' weekly quota too," adding that "it's nearly impossible to do serious work with the current limits, especially on models like Astra" (@eduardo_hsrm ↗).
Tibo then floated a new tier himself: "Would a Pro 50 make sense? Wdyt?" (post ↗). That got more than 1.4 million views and 4,000 replies. "Tibo asking if a Pro 50 makes sense is the easiest yes OpenAI will get all year," replied @ValencianaAbel ↗. To be clear, a Pro 50 is a question on X, not a product.
How it stacks up against Claude
This is now a clear contrast with Anthropic. Claude's pricing page ↗ describes Max ($100) as 5x Pro both per five hours and per week. Ultra ($200) is 10x Pro weekly, with no five-hour limit for verified accounts. We covered why those weekly numbers matter in our report on the Claude Max usage lawsuit.
| At $200/mo | ChatGPT Pro 200 | Claude Ultra |
|---|---|---|
| 5-hour window | None | None (verified accounts) |
| Stated multiple | 20x Plus | 10x Pro weekly |
The two "x" numbers aren't comparable. Plus and Claude Pro are different baselines, and neither company publishes the baseline in tokens. What you can compare is structure. At $100, ChatGPT Pro removes the five-hour window and Claude Max doesn't.
Our take
If Codex is your main agent and you keep hitting the five-hour wall on Plus, Pro 100 is now the obvious upgrade. Removing the session window matters more for long agent runs than the multiplier does. You can start a big refactor at 9 a.m. and not get stopped at lunch.
Be careful in three cases:
Don't buy Pro 500 just for "50x." Unless you need Ultrafast's latency, its 8x burn rate eats the extra headroom.
Screenshot your plan terms. The multipliers live in an X post, not on the pricing page. Anthropic's "5x or 20x" wording ended up in court.
Compare on your own work. Run one week of your normal tasks and watch Settings → Usage. Weekly limits, not the five-hour window, are now the ceiling.
For the agent comparison itself, see our Codex vs Claude Code breakdown and the wider best AI for coding 2026 ranking. If cost is the main issue, start with cheaper Claude Code alternatives.
Sources
Tibo (@thsottiaux) on X ↗, November 8, 2026; Pro 50 question ↗, same day
OpenAI Help Center: Managing usage with GPT-6 Astra in Work and Codex ↗, checked November 9, 2026
OpenAI Help Center: About ChatGPT Pro tiers ↗, checked November 9, 2026
El Mediático: None of ChatGPT's Pro plans has a five-hour limit, OpenAI says ↗, November 8, 2026
Sysnote: Tibo confirms Codex Pro usage ↗, November 8, 2026
Claude pricing ↗, as reported November 8, 2026