GPT-6.2 Codex limits: messages per 5 hours on Plus and Pro
Updated 2026-11-19. From OpenAI's Codex pricing page and Help Center.
OpenAI estimates 17 to 180 local GPT-6.2 Sol messages and 6 to 50 GPT-6.2 Astra messages per 5 hours in Codex on ChatGPT Plus and Standard Business seats. Pro plans have no 5-hour limit right now, only weekly limits. Fast mode halves the count; Ultrafast isn't on Plus.
OpenAI estimates you get 17 to 180 local GPT-6.2 Sol messages and 6 to 50 GPT-6.2 Astra messages every 5 hours in Codex on ChatGPT Plus ($20), per its Codex pricing page, read Nov 19, 2026. Standard Business seats get the same. Pro plans have no 5-hour limit right now, only weekly limits, and OpenAI doesn't publish Pro counts.
Those are GPT-6.2 Codex limits as estimates, not hard caps. OpenAI says the count swings with task size, context, reasoning and tool use, which is why the range is so wide. Both models arrived in Codex and ChatGPT Work on Nov 17, 2026.
GPT-6.2 Codex messages per 5 hours on Plus
Here is every model in OpenAI's table, with the GPT-6.2 pair in bold. The credit columns show why the counts differ: a model that costs more credits per token drains your allowance faster.
| Model | Plus, per 5h | Credits in/out |
|---|---|---|
| GPT-6.2 Astra | 6 to 50 | 200 / 1,000 |
| GPT-6 Astra | 5 to 45 | 250 / 1,250 |
| GPT-6.2 Sol | 17 to 180 | 50 / 250 |
| GPT-6.1 Sol | 15 to 160 | 50 / 250 |
| GPT-6.1 Terra | 35 to 285 | 25 / 125 |
| GPT-6 Sol | 15 to 150 | 50 / 250 |
| GPT-6 Luna | 350 to 3,000 | 2.5 / 12.5 |
OpenAI's estimates of local messages per 5 hours on Plus and Standard Business seats, and Standard-speed credits per 1M input and output tokens, from its Codex pricing page, read Nov 19, 2026. Not fixed limits. Cloud tasks can use more.
GPT-6.2 Sol gets roughly three times as many messages as GPT-6.2 Astra on the same plan. If you only need one number, plan on Sol for most of your day and save Astra for the hard parts. For the per-message cost math in credits, see our Codex credits guide.
Does GPT-6.2 use less of your Codex limit than older models?
Yes, slightly. GPT-6.2 Sol's estimate of 17 to 180 is about an eighth higher than GPT-6.1 Sol's 15 to 160, even though both cost the same 50 input and 250 output credits per million tokens. OpenAI doesn't say why the range moved.
GPT-6.2 Astra is the bigger change. It costs 200 input and 1,000 output credits per million tokens, 20% less than GPT-6 Astra (250 and 1,250), and cached input drops 60%, from 25 to 10 credits. OpenAI's estimate rises from 5 to 45 to 6 to 50 messages per 5 hours. Astra is still the most expensive way to spend a Plus limit.
How Fast and Ultrafast change the count
Fast mode draws your included allowance at 2x the Standard rate on both GPT-6.2 models (2.5x on GPT-6.1 Sol and older models), and it's available on Plus. Ultrafast draws it at 6x and isn't sold on Plus at all, not even with credits, per OpenAI's help article.
| Speed on Plus | GPT-6.2 Sol | GPT-6.2 Astra |
|---|---|---|
| Standard (official) | 17 to 180 | 6 to 50 |
| Fast, 2x (estimate) | About 8 to 90 | About 3 to 25 |
| Ultrafast, 6x | Not on Plus | Not on Plus |
Local messages per 5 hours. The Fast row is our estimate: OpenAI's Standard range divided by its 2x Fast multiplier for included usage.
So a Plus user who leaves Fast on gets about half the messages. That's fine for a quick fix you're watching, and a poor trade for a long background task. Our OpenAI Ultrafast pricing guide has the full speed table and the March 2027 price step.
GPT-6.2 Codex limits by plan
| Plan | Price | 5-hour limit | Ultrafast |
|---|---|---|---|
| Plus | $20 | Yes, see estimates above | No |
| Business (Standard seat) | $25/user | Yes, same as Plus | No |
| Pro 100 | $100 | None right now | No |
| Pro 200 | $200 | None right now | Sol, in Codex only |
| Pro 500 | $500 | None right now | Astra and Sol |
Monthly prices from our Codex plan data (Business is per user, billed monthly). Limits and Ultrafast access from OpenAI's Codex pricing page and Work and Codex help article. Weekly limits can apply on every plan.
Pro 100, 200 and 500 have no 5-hour window right now, only weekly limits, which suits long sessions. OpenAI doesn't say how much more each Pro tier gets in messages; our Pro 100 vs 200 vs 500 comparison covers what it does publish. To use either GPT-6.2 model in the Codex CLI you need version 0.166.0 or later.
Why your Codex limit can still vanish in a day
Agent setups multiply usage. One Codex user told OpenAI's Tibo (@thsottiaux) on X on Nov 19 that a full week's limit went in one day running GPT-6.2 Astra as a coordinator with GPT-6.2 Sol subagents (post). That's a user report, not an OpenAI figure, but the mechanics fit the official rules:
- One shared pool: local messages, cloud tasks and ChatGPT Work all draw from the same allowance, and OpenAI lists no separate allowance for subagents.
- Model choice: every Astra turn costs about four times the credits of a Sol turn at the same token count.
- Speed: Fast doubles the draw, Ultrafast (not on Plus) multiplies it by six.
- Images and context: image generation uses included limits 3 to 5 times faster, and big AGENTS.md files or many MCP servers add tokens to every message.
When the limit resets and what to do when you hit it
The 5-hour window rolls, and weekly limits reset once a week. Your exact reset times are in Settings, then Usage, or type /status in the Codex CLI. We track OpenAI's global resets as they happen on when does Codex reset, with past ones in the Codex reset log.
If you hit the limit mid-task, OpenAI lets the current turn finish, subject to fair use. After that, Plus and Pro users can buy credits, anyone can run local chats with an API key at API rates, or you wait. Our Codex usage limit reached guide walks through each option, and plan vs API shows when paying per token beats a bigger plan.
What changes in December
Starting in December, Plus, Pro and Business allowances will be shared across ChatGPT and Codex and measured in tokens, per OpenAI's Help Center. Instant chat won't count. OpenAI hasn't given a start day or new numbers, so today's 5-hour estimates are the best guide until then. More on that in ChatGPT and Codex will share one usage limit.
What we know and what we don't
| Confirmed by OpenAI | Not published |
|---|---|
| Plus and Business 5-hour estimates for both GPT-6.2 models | Pro message counts or weekly caps |
| Credit rates and Fast and Ultrafast multipliers | How big the weekly limit is on any plan |
| Pro has no 5-hour limit right now | The December switch date or new token numbers |
Astra or Sol on a Plus limit?
Sol, for most work. At 17 to 180 messages per 5 hours it gives you about three times Astra's room. Switch to Astra for the bug that Sol can't crack, then switch back. If you keep running out on Plus, compare a credit top-up with Pro 100, which has no 5-hour window right now. Our GPT-6.2 Astra vs Sol comparison covers the API side.
Sources
- OpenAI Codex pricing: included usage, speed multipliers and credit rates (read Nov 19, 2026)
- OpenAI Help Center: ChatGPT Work and Codex, GPT-6.2 models and Ultrafast (updated Nov 18, read Nov 19, 2026)
- OpenAI Help Center: Understanding ChatGPT and Codex usage limits (read Nov 19, 2026)
- User report on X, reply to @thsottiaux, Nov 19, 2026 (reported, not confirmed)
FAQ
How many GPT-6.2 Sol messages do I get in Codex on Plus?
About 17 to 180 local messages per 5 hours, per OpenAI's Codex pricing page. It's an estimate, not a fixed cap, and a weekly limit may also apply.
How many GPT-6.2 Astra messages do I get in Codex on Plus?
About 6 to 50 local messages per 5 hours, roughly a third of what GPT-6.2 Sol gets. Standard Business seats get the same estimate.
Does GPT-6.2 Sol use more Codex usage than GPT-6.1 Sol?
No. OpenAI estimates 17 to 180 GPT-6.2 Sol messages per 5 hours on Plus against 15 to 160 for GPT-6.1 Sol, and both cost 50 input and 250 output credits per million tokens.
Does ChatGPT Pro have a 5-hour limit for GPT-6.2 in Codex?
No. OpenAI says Pro plans currently have no five-hour limit, though weekly limits may apply. It doesn't publish Pro message counts.
Why does my Codex limit run out so fast on GPT-6.2?
Usually model, speed and agents. Astra costs about four times Sol's credits per token, Fast doubles the draw, Ultrafast multiplies it by six, and subagents share the same pool.
Will GPT-6.2 Codex limits change in December?
Yes, in how they're counted. From December, Plus, Pro and Business allowances will be shared across ChatGPT and Codex and measured in tokens. OpenAI hasn't published a start day or new numbers.
Related
- When does Codex reset? (live)
- Codex usage limits explained
- Codex usage limit reached: what to do
- Codex credits: prices and rates
- OpenAI Ultrafast pricing: GPT-6.2 API and Codex costs
- ChatGPT and Codex will share one usage limit
- GPT-6.2 Astra vs GPT-6.2 Sol
- ChatGPT Pro 100 vs 200 vs 500
- Hit the Codex weekly limit? The cheapest API fallback
- Codex pricing and changes
More guides
- GPT-6.1 Terra vs GPT-6.1 Sol: which should you use?
- GPT-6 Astra vs GPT-6.1 Sol
- GPT-6 Astra vs Claude Opus 5.5: cost, coding and usage limits
- Claude Fable 5.2 and Mythos 5.2: prices, limits and what changed
- Claude Fable 5.2 vs Mythos 5.2: which one should you use?
- Is Claude Pro enough for Fable 5.2? Fable and Mythos limits by plan
Not sure a plan beats the API for you? Try the plan vs API cost calculator.