ChatGPT, Claude, Gemini and Copilot plans: AI usage limits use different units

The central comparison problem is not simply that the allowances differ. Each cited vendor page describes usage through a different measurement system. Claude uses five-hour session windows and plan-to-plan multipliers (https://claude.com/pricing). Google’s UK Gemini page expresses paid access as multiples of the no-cost tier or AI Pro, without supplying a common absolute message count (https://gemini.google/gb/subscriptions/?hl=en-GB). Microsoft’s consumer Copilot page uses qualitative phrases such as higher, extensive and highest usage for select features, while stating that limits apply (https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/individuals). The supplied OpenAI page documents local messages per five-hour period, shared local and cloud allowances, and possible weekly limits—but specifically in the context of Codex within ChatGPT plans (https://learn.chatgpt.com/docs/pricing). These units cannot responsibly be collapsed into one messages-per-month leaderboard. A useful comparison must preserve the vendor’s exact scope, window and account rules.

Claude: five-hour windows and relative plan multipliers

Anthropic’s page lists Claude Free at $0 and names Sonnet and Haiku among its models. It also lists chat across web, mobile and desktop, plus capabilities including web search, memory, file creation, code execution and extended thinking (https://claude.com/pricing). Claude Pro is shown at $17 per month when billed annually or $20 monthly. The page describes Pro as providing at least five times more usage per five-hour session than Free and adds Claude Code, Claude Cowork, Claude Design, unlimited projects and access to Opus (https://claude.com/pricing). Claude Max starts at $100 per month and is described as providing five or twenty times more usage than Pro per five-hour session, with higher output limits for all tasks (https://claude.com/pricing). The practical detail is the denominator: these are session-window comparisons, not guaranteed monthly message totals. A buyer should therefore examine when their intensive work happens. Occasional long sessions and evenly distributed daily use may encounter the same window in very different ways.

Gemini: UK prices paired with tier multipliers

Google’s cited page is explicitly a UK subscription page, so its GBP prices should not be generalized to other regions. It shows a no-cost Gemini tier at £0 per month with a Google Account, including access to 3.6 Flash, varying access to 3.1 Pro and 15 GB of storage (https://gemini.google/gb/subscriptions/?hl=en-GB). Google AI Plus is listed at £4.49 per month with twice the usage limits of the no-cost tier. Google AI Pro is listed at £18.99 per month with four times the no-cost limits, 1,000 Google Flow credits and 5 TB of storage (https://gemini.google/gb/subscriptions/?hl=en-GB). Google AI Ultra is shown as starting at £79.99 per month, with another £189.99 tier displayed, and the page describes five or twenty times higher usage than AI Pro (https://gemini.google/gb/subscriptions/?hl=en-GB). Multipliers communicate relative position, but without an absolute baseline they do not establish a predictable task count for an individual workflow.

Copilot: qualitative limits and subscription-owner restrictions

Microsoft’s individual pricing page bundles Copilot with Microsoft 365 subscriptions. It lists Personal at $9.99 per month or $99.99 per year, Family at $12.99 per month or $129.99 per year, and Premium at $19.99 per month or $199.99 per year (https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/individuals). Personal is described as offering higher usage than free for select Copilot features, while Premium is associated with extensive and highest usage language. The page does not provide numeric credit amounts and states that usage limits apply (https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/individuals). It also says AI features are for the subscription owner and cannot be shared, including in the Family plan (https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/individuals). Microsoft Support says free Copilot is available on the web, in apps and through Edge, while eligible Microsoft 365 plans add Copilot inside Word, Excel, PowerPoint, Outlook and OneNote for the owner (https://support.microsoft.com/en-us/microsoft-365-copilot/what-s-the-difference-between-microsoft-copilot-free-and-copilot-in-microsoft-365).

ChatGPT: the published limits describe Codex, not every chat

OpenAI’s pricing page lists ChatGPT Free at $0 per month, Go at $8, Plus at $20, Pro from $100, and Business at $20 per user per month when billed annually for at least two users, with Enterprise and Edu handled through sales (https://learn.chatgpt.com/docs/pricing). Read the usage section carefully, though: the numbers it gives describe Codex, OpenAI’s coding agent, inside those plans. Free can try Codex on quick coding tasks, and allowances are counted as local messages per five-hour period (https://learn.chatgpt.com/docs/pricing). Local messages and cloud chats draw on the same allowance, and weekly limits may apply on top. The page gives model-specific ranges for some plans, Plus among them, but those are Codex figures and should not be read as a general ChatGPT message quota (https://learn.chatgpt.com/docs/pricing). That distinction matters for a buyer: a limit published for a coding feature says nothing about how much chat, research or image work a plan allows. For those, look for a separate statement from OpenAI before comparing it with the other vendors here.

A practical buying method that avoids false precision

Start by naming one job rather than choosing a brand in the abstract: coding sessions, document work, research, general questions or another repeatable task. Next, copy the vendor’s exact limit language into a simple table. Keep five-hour windows, free-tier multipliers and qualitative labels in separate columns; do not convert them into invented monthly totals. Claude’s page supports a session-window comparison (https://claude.com/pricing), while the UK Gemini page supports relative tier multipliers (https://gemini.google/gb/subscriptions/?hl=en-GB). Microsoft supports feature and ownership comparisons but not a numeric consumer-limit calculation on the cited pages (https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/individuals; https://support.microsoft.com/en-us/microsoft-365-copilot/what-s-the-difference-between-microsoft-copilot-free-and-copilot-in-microsoft-365). The OpenAI source supports Codex-specific analysis only (https://learn.chatgpt.com/docs/pricing). Finally, keep a one-week diary of tasks, interruptions and apparent resets. That record will not reveal hidden limits, but it can show whether the lowest suitable tier fits your actual work before you consider an upgrade.

Sources

Narration and illustrations in the companion video were generated with AI. Every claim above was checked against the linked pages on 2026-09-08. Article text expanded on 2026-09-08.

Spotted an error? Report it here. We fix it and list it, dated, on the Corrections page →