Your ChatGPT usage limit, tracked live
ChatGPT limits you per model, and shows no counter until you hit the wall. tokenkarma surfaces the live count and reset time for each model, so you never get stopped mid-task again.
The main consumer window is rolling: on Plus and Go you can send up to 160 GPT-5.5 messages every 3 hours, per OpenAI's official documentation, and capacity returns as the window rolls. Hit the cap and ChatGPT switches to the mini model until the limit resets, rather than blocking you. Free runs on a tighter five-hour window whose limits OpenAI describes as dynamic.
How ChatGPT's usage limits work
Unlike Claude's single shared budget, ChatGPT meters each model on its own allowance. You can run out on one model while another still answers, which is why the wall often feels arbitrary. Three mechanics matter.
The published per-model window
For GPT-5.5, the everyday model, OpenAI publishes one concrete consumer number: "ChatGPT Plus and ChatGPT Go users can send up to 160 messages with GPT-5.5 every 3 hours." When you reach it, "chats will switch to the mini version of the model until the limit resets". On Go, manually selected Thinking has its own counter: up to 10 messages every 5 hours, and automatic switching into Thinking does not count against it.
Dynamic limits on the lower tiers
Free-tier access runs on a five-hour window and is explicitly not fixed: limits "can be dynamic and may vary based on factors that include market, system conditions, abuse-prevention guardrails, and individual usage" (same source). The Go plan page carries similar wording. In practice: your ceiling moves with demand, and OpenAI does not owe you the same number two days in a row.
Separate counters for heavy reasoning
The heaviest models are metered separately. On Business plans, OpenAI publishes the counters: GPT-5.5 Instant is "virtually unlimited", GPT-5.5 Thinking is capped at 3,000 requests per week, and GPT-5.5 Pro at 15 requests per month, with older models on their own separate counters. On Pro plans, the Pro-model allowance differs by tier and is not published as a number; when you reach it, "ChatGPT will use another available model until your Pro-model allowance resets" (source).
ChatGPT plans and their limits, side by side
OpenAI's current consumer lineup, with what each plan officially documents about limits. Every cell links to the page it comes from; prices are OpenAI's published US prices as of June 2026.
| Plan | GPT-5.5 messages (official) | Heavy reasoning | Source |
|---|---|---|---|
| Free ($0) | "Limited" access in a 5-hour window, dynamic, no published count | Not included | GPT-5.5 limits |
| Go ($8/mo) | Up to 160 every 3 hours | 10 manual Thinking messages per 5 hours | GPT-5.5 limits |
| Plus ($20/mo) | Up to 160 every 3 hours | Thinking included; no published weekly count | Plus plan |
| Pro ($100 or $200/mo) | "Unlimited" GPT-5 access, subject to abuse guardrails | Pro-model allowance by tier: $100 = 5x Plus usage, $200 = 20x | Pro tiers |
| Business ($20-25/user/mo) | GPT-5.5 Instant "virtually unlimited" | 3,000 Thinking/week, 15 GPT-5.5 Pro/month | Business limits |
Current prices and plan contents: OpenAI's pricing page. Deep research and agent-mode quotas are tiered ("limited" on Free and Go, "expanded" on Plus, "maximum" on Pro) but OpenAI publishes no per-plan counts; fixed monthly allowances reset every 30 days from first use (deep research FAQ).
Reset cheat sheet
| Limit | Window | What happens at the cap |
|---|---|---|
| GPT-5.5 (Plus, Go) | 3-hour rolling | Switches to the mini model until reset |
| GPT-5.5 (Free) | 5-hour window, dynamic | Notice + upgrade prompt; GPTs blocked until the stated time |
| Thinking (Go, manual) | 5-hour rolling | Counter shown; automatic Thinking does not count |
| Thinking (Business) | Weekly | 3,000 requests per week per seat |
| GPT-5.5 Pro (Business) | Monthly | 15 requests per month per seat |
| Pro-model allowance (Pro plans) | Not published | Falls back to another model until the allowance resets |
One more documented reset path: OpenAI confirms that Free-tier rate limits reset immediately when you upgrade to a paid plan.
Tracking beats guessing
See your ChatGPT windows live
Per-model usage, reset times, alerts before the wall. Join the early access.
How to see your ChatGPT usage
ChatGPT gives you no day-to-day view of your remaining messages. The first sign of trouble is usually the wall itself: the silent switch to the mini model, or a cooldown notice. tokenkarma closes that gap. For each model it shows:
- the live count used and how much is left,
- the exact time the per-model window resets,
- the weekly and monthly counters where they exist,
- an alert before you hit the wall, not after.
For the wider background on what changed in 2026, read the OpenAI IPO filing breakdown and the hidden rate limits of every major AI API.
Track ChatGPT next to every other AI you pay for
Most people who pay for ChatGPT also pay for Claude, Gemini, Cursor or Perplexity, each with its own limits and its own blind spots. tokenkarma puts them all in one view, so you can move work to whichever model still has room. See the full AI usage tracker, the side-by-side limit comparison, or how Claude's limits differ from ChatGPT's.
Frequently asked questions
What are ChatGPT’s usage limits?
ChatGPT meters usage per model on short rolling windows. The one number OpenAI publishes for consumer plans: Plus and Go users "can send up to 160 messages with GPT-5.5 every 3 hours" (official source). Free runs on a tighter, dynamic five-hour window with no published count. Heavier reasoning use has its own counters: on Business, GPT-5.5 Thinking is capped at 3,000 requests per week and GPT-5.5 Pro at 15 per month.
How many ChatGPT messages can I send per day?
There is no daily number. The published consumer limit is per window: up to 160 GPT-5.5 messages every 3 hours on Plus and Go. Free-tier limits are explicitly dynamic: OpenAI says they "may vary based on factors that include market, system conditions, abuse-prevention guardrails, and individual usage". So two Free users can see different ceilings on the same day.
When does my ChatGPT limit reset?
The per-model window is rolling: capacity returns about 3 hours after the messages that used it (5 hours on Free). When you hit the GPT-5.5 cap, OpenAI states that "chats will switch to the mini version of the model until the limit resets", so you are downgraded rather than blocked. Upgrading also resets you: OpenAI confirms Free-tier rate limits reset when you move to a paid plan.
Why did I hit my ChatGPT limit with no warning?
ChatGPT shows no running counter before you hit the wall: the first signal is the switch to the mini model or a cooldown message. Limits also flex with demand on Free and Go: OpenAI says limits "may vary based on system conditions to ensure a smooth experience for all users". That blind spot is exactly what tokenkarma fills, surfacing the live count and reset per model before you are stopped.
What is the difference between Plus, Pro and Business limits?
Plus gets the published 160 messages per 3 hours on GPT-5.5. Pro comes in two tiers: per OpenAI, "Pro $100 unlocks 5x higher usage than Plus, while Pro $200 unlocks 20x usage than Plus", with unlimited access to GPT-5 models "subject to abuse guardrails". Business publishes its own counters: virtually unlimited GPT-5.5 Instant, 3,000 Thinking requests per week and 15 GPT-5.5 Pro requests per month per seat. Sources are linked in the table above.
Can I track my ChatGPT usage in real time?
Yes. tokenkarma reads your live ChatGPT usage and shows how much of each per-model window you have used and when it resets, with an alert before you hit the wall. It runs as a Mac menu-bar app, a web dashboard and a Chrome extension, alongside your other AI tools.
Is tokenkarma affiliated with OpenAI?
No. tokenkarma is independent and not affiliated with OpenAI. The limits described here reflect OpenAI’s published documentation as of June 2026, every number links to its official source, and they change often.
Stop guessing your ChatGPT limit
tokenkarma tracks your per-model ChatGPT windows live, with reset times and alerts before you hit a wall.
Free to start. Works in your browser and on your Mac.
We already ship a free Chrome extension for Claude limits, claudeKarma, used by 12,000+ people: see it on the Chrome Web Store.
Not affiliated with OpenAI. Limit mechanics reflect OpenAI's published documentation as of June 2026, linked above, and can change frequently.