People search "chatgpt limit" and "claude limit" hoping for a number. Most of the time the official pages don't give one. This is what Anthropic and OpenAI actually say, plan by plan, as of 10 October 2026, and where the numbers stop. We read every page cited here on that date, and the date matters: these pages change often.
Why a limit runs out faster than expected, and what actually uses it up, is in why Claude hits its usage limit so fast.
How Claude counts usage
Every paid Claude plan has two limits. A session limit resets every five hours. A weekly limit applies across all models and resets at a fixed time assigned to your account, which you can see in Settings > Usage. The Free plan has the five-hour session limit, and Anthropic says the number of messages "will vary based on demand".
Anthropic doesn't publish a message count or a token count for any plan. It publishes multiples. Max 5x gets five times the Pro plan's per-session allowance and Max 20x gets twenty times. On Team, a Standard seat gets 1.25 times Pro and a Premium seat 6.25 times, and the limits are per member, not shared by the team.
One detail matters if you also code: usage in claude.ai, Claude Code and Claude Desktop counts toward the same limit. A long Claude Code session in the morning leaves less for chat in the afternoon.
How ChatGPT counts usage
OpenAI splits Chat from Work and Codex. In Chat, Free and Go users have "unlimited everyday text chats, subject to abuse-prevention safeguards". File uploads, image generation, voice, data analysis, the thinking levels and the Pro models have their own limits, which depend on the plan.
Work and Codex share a separate allowance. Depending on the plan, it runs on a five-hour window and a weekly window, and you need allowance left in both to keep going. OpenAI says plainly that this allowance "does not correspond to a fixed number of messages or tasks", but it does give estimated ranges per model for Plus and Business Standard.
The table
| Plan | What the official page says | Window | Context window |
|---|---|---|---|
| Claude Free | Messages vary with demand | 5 hours | Not stated |
| Claude Pro | More per session than Free, no number | 5 hours + weekly | Paid plans: up to 1M tokens on the newest models, 500K or 200K on others |
| Claude Max 5x / 20x | 5× / 20× Pro per session | 5 hours + weekly | As above |
| Claude Team | Standard 1.25×, Premium 6.25× Pro per session, per member | 5 hours + weekly | 200K (Team plan page) |
| ChatGPT Free | Unlimited everyday text chats; tools limited separately | Reset time shown in the app | Not in the help pages |
| ChatGPT Go | Same text chats; higher tool limits than Free | Reset time shown in the app | "Larger" than Free, no number |
| ChatGPT Plus | May include message caps, "especially during high demand". In Work and Codex: GPT-6 Sol 15–150 local messages per 5 hours (estimate) | 5 hours + weekly in Work and Codex | Not in the help pages |
| ChatGPT Pro 100 / 200 / 500 | GPT-6 Pro has limits, not published. No five-hour limit in Work and Codex | Weekly | Not in the help pages |
| ChatGPT Business | Standard: 15 Pro messages a month. Premium: 50 a week, 5× Standard's Work and Codex usage, no five-hour limit | 5 hours (Standard) | 54K Instant, 256K Reasoning |
| ChatGPT Enterprise | Credits and rate card | Set by contract | 128K Instant, 256K Reasoning |
Checked on 10 October 2026 on support.claude.com, help.openai.com and openai.com. Limits change often: several of these pages had been updated in the previous few days, and on Pro 200 new subscriptions already get a lower allowance; only accounts with an active Pro 200 between 22 September and 29 September 2026 keep the old one, until 29 October 2026. Before relying on a figure, open Settings > Usage in your own account. Both companies show your remaining allowance and reset times there.
Where the table says "no number", there really isn't one. Anthropic doesn't publish how many messages a Pro session holds, and OpenAI doesn't publish a fixed message cap for Plus in Chat. The figures you see on forums come from people measuring their own accounts.
Why the same plan runs out at different speeds
Both companies say the same thing in different words: the limit depends on how much text the model has to process. Anthropic lists conversation length, attachment size, the model, the effort level, and tools and connectors, which it calls "token-intensive". Project instructions and files take up context space too. OpenAI's Pro page lists "the amount of context processed, reasoning, tool use, information retrieval, caching".
Claude Code's documentation explains the mechanism. The whole conversation is sent with every request, so a one-line question at the end of a long session still draws usage for everything before it. Cached content counts less against your limits when it's reused. Anthropic doesn't say how much less, so we won't guess.
In practice, what you carry into each message counts more than how many messages you send. If you open every chat by pasting the same project brief, the same decisions and the same client notes, you spend part of your allowance on text the model has already seen many times.
What a memory outside the chat changes
We measured this on the Cobrain project in our own brain, 134 notes, with a Claude token counter built from 258 real Claude Code conversations and checked against OpenAI's. On these notes (Italian plus Markdown) Claude counts about 2 characters per token, about 1.58 times OpenAI's count for the same text.
Giving the AI the files the project marks as always needed (the overview, the index, the decisions) in full at every chat costs about 69,400 tokens. Cobrain's opening summary of the same project is 4,700 tokens. Cobrain also has a fixed cost when it's connected: the definitions of its tools take about 5,200 tokens. With that included, the first message is 69,400 against roughly 9,900, which is 7 times less.
Over whole working chats, 35 real ones on the same project with every note read included, the median is 2.9 times less. Counting that the model rereads the context at every reply, it's 3.6 times less. In 4 of the 35 chats Cobrain cost more than the "without" case. Those chats opened a lot of notes, and without Cobrain that material would have had to be given anyway. The full method is on the context costs page.
Two limits on this. Any MCP connector adds tokens, Cobrain included, because its tool definitions travel with the conversation. In Claude Code they're deferred by default, so only tool names and server instructions load until a tool is used. In the chat apps we don't know how each one loads them. And a memory doesn't raise your plan's limit. It only reduces how much you send.
Where Cobrain fits
Cobrain keeps your notes in real folders that Claude, ChatGPT, Claude Code and Codex read and write through a connector. A chat opens with the brain's rules, the project overview and the latest decisions; after that the AI searches and opens only the notes it needs. The guide for Claude and the guide for ChatGPT take a few minutes, and the free plan holds up to 1,000 notes.
Quick answers
What is the ChatGPT Plus message limit? OpenAI doesn't publish a fixed number for Chat. Its Plus page says there may be message caps, especially during high demand. For Work and Codex it gives estimates per five-hour window, such as 15–150 local messages with GPT-6 Sol.
Is there a ChatGPT token limit per day? OpenAI doesn't publish a daily token figure. Free and Go have unlimited everyday text chats. The limits apply to tools, thinking levels, Pro models, and Work and Codex, which use five-hour and weekly windows.
When do Claude usage limits reset? The session limit resets every five hours. On paid plans the weekly limit resets at a fixed time assigned to your account, shown in Settings > Usage.
Does Claude Code share usage limits with Claude? Yes. Usage in claude.ai, Claude Code and Claude Desktop counts toward the same limits.
Claude usage limits vs Codex: what's the difference? On most plans both use a five-hour window and a weekly window. Anthropic publishes multiples of the Pro plan. OpenAI publishes estimated message ranges per model. Neither publishes a guaranteed count.
Why do I hit the Claude limit after only a few messages? Usually because of what each message carries: a long conversation, large attachments, connectors, a high effort level. Starting a new chat, and giving the AI only the notes it needs, makes each message lighter.
