ChatGPT vs Claude vs Gemini Usage Limits Compared (2026)

I hit Claude Pro’s 5-hour limit often. I’ve never hit the weekly one. All three $20 plans now run on those two clocks, and each company counts differently.

OpenAI, Anthropic, and Google all sell consumer plans around $20, $100, and $200. Google adds a $4.99 tier, OpenAI adds $8 and $500 tiers, and Anthropic has nothing below $20.

ChatGPT is about to change. Today Chat has its own limits, while Work, Codex, and the Office add-ins share a separate allowance. Starting in December, OpenAI says Plus, Pro, and Business will share one token-based allowance across ChatGPT and Codex, with Instant chat exempt. It hasn’t given a date or published allowance sizes. Claude already works this way: chat and Claude Code draw from one pool. Google’s Gemini app runs on compute-based limits and publishes no prompt count at all.

Here’s how each price tier compares, how much each model costs you, and when the limits reset.

ChatGPT vs Claude vs Gemini Limits at a Glance

• Under $10: Google AI Plus ($4.99) doubles the free Gemini limits but doesn’t include Gemini 4 Argon. ChatGPT Go ($8) gets you unlimited everyday chat, though it may include ads. Anthropic has no plan here.
• $20: All three have a 5-hour window and a weekly limit. Only ChatGPT Plus prints numbers: OpenAI estimates 15–160 local messages per 5 hours on GPT-6.1 Sol. SemiAnalysis measured ~2.9B tokens a month on Claude Pro against ~1.0B on Plus in an October coding-agent workload. Google AI Pro ($19.99) adds Argon, video credits, and 5 TB of storage.
• $100: ChatGPT Pro 100 drops the 5-hour limit in Work and Codex and gives you 10x Plus. Claude Max gives you 5x Pro and includes Fable 5.1. Google AI Ultra ($99.99) gives you 5x AI Pro.
• $200: ChatGPT Pro 200 gives you 20x Plus. Claude Ultra gives you 10x Pro every week, with no 5-hour limit on verified accounts. Google AI Ultra ($199.99) gives you 20x AI Pro.
• $500: Only OpenAI sells one: ChatGPT Pro 500, at 50x Plus with Ultrafast mode.

Prices and limits come from OpenAI, Anthropic, and Google pricing and help pages, re-checked 2026-11-06. OpenAI’s current pages don’t print its Pro multiples, so How this was checked shows where they come from.

Budget Plans: Google AI Plus ($4.99) vs. ChatGPT Go ($8)

Anthropic doesn’t sell anything between Claude Free and Claude Pro ($20), so a sub-$10 upgrade means Google AI Plus or ChatGPT Go.

Neither budget plan gets a flagship: Google reserves Argon for AI Pro and Ultra, and OpenAI keeps Free and Go on Luna in Chat. The bigger daily difference is that Go’s everyday chat stays unlimited, subject to abuse guardrails, while AI Plus runs on a limit that shrinks with long chats and heavy features. For $3 less, though, AI Plus adds the Flow credits and storage.

$20 Plans: ChatGPT Plus vs. Claude Pro vs. Google AI Pro

Only Claude pools web chat and coding today. ChatGPT joins it in December, and Google gives Antigravity its own 5-hour and weekly limits.

*Monthly tokens and API-list value from SemiAnalysis’s October 5, 2026 coding-plan study: an agent workload, with Claude on Opus 5.5 and ChatGPT on GPT-6.1 Sol. API-list value is what the same tokens would cost on the API.

How Many Messages Does ChatGPT Plus Allow?

It depends on the model, and OpenAI gives a range. Plus gets Medium and High thinking in Chat, and everyday Instant chat stays unlimited. In Work and Codex, OpenAI publishes estimated local messages per 5 hours, plus a token rate for each model. Those estimates aren’t fixed limits; cloud tasks can cost more, and a weekly limit may also apply.

That’s a 100x spread in token cost between Luna and Astra. Other tools draw down the same allowance:

• Images: Image generation uses your limit 3x to 5x faster than a similar turn without it.
• Office add-ins: A typical Excel or Sheets task uses 5–20 credits per message, and PowerPoint uses 10–50.
• Desktop Voice: Voice in Work and Codex costs $0.05 of usage per minute, or 1.25 credits a minute on purchased credits, plus the model’s tokens.
• Fast mode: It burns your included allowance at 2.5x Standard speed, or 2x on purchased credits.

Uploads have their own cap: 80 files every 3 hours on paid plans. Regular Chat Voice has separate limits too.

How Do Claude Usage Limits Work?

Claude has a 5-hour session limit and a weekly limit, and they cover everything: web, desktop, mobile, and Claude Code. Anthropic publishes no message count, because usage depends on conversation length, attachments, model, and effort level.

On SemiAnalysis’s coding workload, Claude Pro got about 2.9x the tokens of ChatGPT Plus, which works out to about 5.6x the API-list value. On my own Claude Pro account, normal writing, research, and coding hit the 5-hour limit often and have never hit the weekly limit. For the full coding breakdown, see our Claude Pro vs ChatGPT Plus usage limits comparison.

How Do Google AI Plan Limits Work?

Google publishes multipliers—no counts. AI Plus gets 2x the free limits, AI Pro gets 4x, and AI Ultra gets 5x or 20x AI Pro. Your usage depends on the prompt, the model, the feature, and how long the chat runs, and Deep Research and video cost the most.

Google AI Pro’s 1M-token context window comes with a catch in Antigravity: it compacts chats at about 256K tokens, and Google staff call that threshold a fixed default you can’t change yet.

If you want a hard Gemini 4 Argon number for coding, OpenCode Go ($10/month) publishes 440 Argon requests per 5 hours during Argon’s 50% discount through November 30, 2026. Our Gemini 4 Argon usage limits guide goes deeper on Antigravity.

Antigravity is the odd one out. AI Pro and Ultra also get Claude Sonnet 5.5 and Opus 5.5 inside it, on a separate limit from the Gemini models. Ultra gets 5x or 20x Pro’s tokens there too.

$100 Plans: ChatGPT Pro 100 vs. Claude Max vs. Google AI Ultra 5x

Google and Anthropic sell 5x their $20 plans here. OpenAI’s October 29 doubling puts Pro 100 at 10x Plus in Work and Codex.

• Claude Max includes Fable 5.1: On $20 Pro, Fable 5.1 ($10/$50 per 1M API tokens) runs on pay-as-you-go usage credits from the first message. On Max it’s included for up to 50% of your weekly limit, and after that you buy credits or switch models.
• Argon alone isn’t a reason to leave AI Pro: Deep Think was the old Ultra draw, and Google says Argon is replacing it. Ultra buys you higher limits and bigger extras. (I’m on Ultra 5x myself, for what it’s worth.)

The API credits on ChatGPT Pro, Claude Max, and Claude Ultra don’t buy more app usage. OpenAI’s don’t raise your ChatGPT or signed-in Codex limits, and Anthropic’s can’t be spent in Claude Code or the Claude apps. They’re for apps you build on the API, and both expire at the end of the billing cycle.

$200 Plans: ChatGPT Pro 200 vs. Claude Ultra vs. Google AI Ultra 20x (Plus ChatGPT Pro 500)

At $200, OpenAI and Anthropic both drop the 5-hour limit, and their weekly allowances split: 20x Plus on Pro 200, 10x Pro on Claude Ultra. Those are multiples of different base plans, so 20x isn’t twice as much as 10x. Google keeps the 5-hour window and sells 20x AI Pro.

• Claude Ultra is the renamed Max 20x: It has the same $200 price and no 5-hour limit once Anthropic verifies your account. Its weekly limit didn’t change: Anthropic’s product lead says it was already 10x Pro, and only the old 5-hour limit was 20x. If you see a Current session bar in Settings → Usage, you’re still on a temporary limit.
• ChatGPT Pro 200 didn’t get its cut: On October 29, OpenAI canceled the reduction it announced on September 29 and raised subscribers on the lower allowance to the higher one. New subscriptions get the same allowance.
• ChatGPT Pro 500 ($500/month): OpenAI is the only one of the three with a $500 personal plan. It has a 50x Plus allowance, no 5-hour limit in Work and Codex, $500 a month in API credits, and Ultrafast on GPT-6 Astra and GPT-6.1 Sol. Ultrafast burns your included usage at 8x Standard speed, or 6x on purchased credits.

When Do the Limits Reset?

• 5-hour limits: On Claude, the timer starts with your first message and the limit resets in full five hours later. ChatGPT works the same way in Work and Codex on Plus. Google says Gemini limits refresh every 5 hours until you reach your weekly limit.
• Weekly limits: Anthropic resets your weekly limit at a fixed day and time assigned to your account, and it’s shown in Settings → Usage. OpenAI and Google publish no fixed reset day. ChatGPT shows your reset times in Settings → Usage and through /status in the Codex CLI, and Gemini shows yours under Settings → Usage Limits.
• Paying past the limit: Claude Pro, Max, and Ultra let you turn on prepaid usage credits at API rates ($10/$50 per 1M tokens on Fable 5.1, $4/$20 on Opus 5.5, $2/$10 on Sonnet 5.5). ChatGPT Plus and Pro can buy credits, valid for 12 months, for Work, Codex, and the Office add-ins, and eligible accounts can buy an instant reset. Google AI subscribers can continue on Flash-Lite.
• Multiple accounts: OpenAI’s card and device limits can block a new Go, Plus, or Pro purchase, or sign you out, when too many paid accounts sit on one card or device.

Which Plan Fits How You Work

• Under $10: Pick Google AI Plus for Flow credits and storage, or ChatGPT Go if you mostly chat. If you want Argon, you need AI Pro.
• At $20: Pick Claude Pro if most of your usage is long documents and heavy coding in Claude Code. Pick ChatGPT Plus if you want published estimates, image generation, and Codex. Pick Google AI Pro if you want Argon and the biggest media and storage bundle.
• At $100–$500: Pick ChatGPT Pro 100 to drop the 5-hour limit without paying $200, or Pro 500 if you need Ultrafast. Pick Claude Max or Ultra if your work depends on Fable 5.1 and large Claude Code repositories. Pick Google AI Ultra for high-volume Flow video, 20–30 TB of storage, or higher Antigravity limits.

If you want one subscription for everyday non-coding tasks, see our best AI subscription for personal use guide, or run your own usage through the AI subscription finder.

What Changed This Month

• November 2026 (OpenAI): OpenAI’s help center added the December shared allowance.
• October 30, 2026 (Anthropic): Max 5x became Max and Max 20x became Claude Ultra at the same prices. Verified Ultra accounts lost the 5-hour limit.
• October 29, 2026 (OpenAI): OpenAI doubled the Work and Codex allowance on Pro 100 and Pro 500, canceled the Pro 200 reduction, and left Plus unchanged.
• October 28, 2026 (OpenAI): Eligible Pro subscribers can claim $100, $200, or $500 in monthly API credits.
• October 26, 2026 (OpenAI): GPT-6.1 Terra arrived in Work and Codex on Plus and above.
• October 13, 2026 (Google): Gemini 4 Argon launched for AI Pro and Ultra in the Gemini app, at $2/$10 per million API tokens through November 30 and $4/$20 after that.

How This Was Checked

I re-checked everything on 2026-11-06, and here’s where it came from.

• OpenAI: pricing, ChatGPT and Codex usage limits, GPT-6 and other models, Work and Codex, Codex pricing, credits, Pro tiers, card and device limits, API credits, file uploads, and the release notes.
• Anthropic: pricing, API pricing, Pro, Max, Ultra, usage and length limits, Claude Code on your plan, usage credits, and Fable on your plan.
• Google: Google AI plans, Gemini Apps limits, Deep Think, and the Argon launch post.
• ChatGPT Pro multiples: OpenAI’s current pages don’t print them. Its release notes launched Pro 100 at a 5x standard allowance, and it doubled that on October 29, which makes 10x. Pro 200 stayed at 20x when OpenAI canceled the cut to 10x, and OpenAI’s Codex lead confirmed 50x for Pro 500.
• The token numbers: Claude Pro’s 2.9B tokens ($1,178) and ChatGPT Plus’s 1.0B tokens ($211) come from SemiAnalysis’s October 5, 2026 coding-plan study. That study predates OpenAI’s October 29 Pro increase, and Plus is unchanged since then. SemiAnalysis’s newest public update, on October 22, found two Max 20x accounts with different limits: one had a 19% bigger 5-hour window and a 20% smaller week. Anthropic hasn’t confirmed a test. I found no newer full study, and nobody has published a measurement of Gemini’s limits.
• What I’ve used: I pay for and use Claude Pro daily across Claude Desktop and Claude Code CLI, and I’ve used Max 5x. I’m on Google AI Ultra 5x now and have used AI Pro, including in Antigravity. I’ve used ChatGPT Plus and two Business seats, but never ChatGPT Pro. ChatGPT Go and Pro, Claude Ultra, Google AI Plus, and Ultra 20x are compared from published documentation. Prices and limits move, so confirm at checkout.

If you want published per-model message estimates on a $20 bill, ChatGPT Plus is the only one of the three that prints them. Claude Pro holds about three times Plus’s tokens on SemiAnalysis’s coding workload, and Google AI Pro gives you the numbers for everything—except Gemini itself.