AI models · Updated 2026-10-31

How usage limits work on Claude Max and ChatGPT Pro

Understand what the subscription tiers buy, why there is no fixed message count, and how to check the reset before a work session.

The short answer

Claude Max and ChatGPT Pro sell larger, still-limited allowances; neither is a promise of unlimited work. A fixed message count is misleading because the model, prompt length, reasoning, files, tools, speed mode, and task all change how quickly an allowance is consumed. Check the usage screen for your account before planning around a reset.

Claude Max

Anthropic currently lists Max 5x at $100 per month and Max 20x at $200 per month, describing those tiers as five or twenty times Pro capacity. The comparison is a plan-level relationship, not a promise of a certain number of messages in every five-hour window. Anthropic says message length, conversation length, attachments, tool use, model choice, and features such as research affect usage. Reusing cached project content can reduce repeated context processing.

A practical check is to open Claude’s usage display during a normal week and note which model and tools consume the allowance. A long coding session with a large repository context may use more than many short questions. If usage runs out sooner than expected, shorten the context, split unrelated tasks into separate chats, or use a less expensive model for the parts that do not need the strongest one. Confirm which Max tier and current limit apply to your own account; limits can change.

ChatGPT Pro

OpenAI currently lists Pro 100, Pro 200, and Pro 500 at those monthly prices. The higher tiers provide larger included usage; Pro 500 adds Ultrafast access for certain models. Plan usage depends on model, context, reasoning, tools, retrieval, and speed mode. For example, OpenAI says Fast mode consumes the included allowance faster than Standard, and Ultrafast uses it faster still. Eligible paid Pro and Business subscribers can also claim separate credits for paid API Platform usage in a linked API organization. Those credits expire at the billing-cycle end and do not raise ChatGPT or signed-in Codex limits; Codex use through API keys is excluded. They are not a way to continue a capped chat or Codex session. See the API-credit eligibility and billing rules.

Limits may be separate by model or product surface. OpenAI documents GPT-6 Pro chat limits separately from Work and Codex allowances, and some Pro tiers may switch to another model after a limit. The usage page shows the reset when available. A Pro subscription remains active when an individual model limit is reached; the model may simply be unavailable until it resets.

Do not compare the multipliers directly

“5x Pro” and a dollar price do not measure the same thing, and providers do not publish an apples-to-apples unit of work. A long context, high reasoning level, tool-heavy coding task, or fast mode can change the effective work completed per dollar. Compare the plans on a week of your own tasks: note the model, task type, remaining allowance, and reset time at the start and end of each session.

A reliable planning routine

Before a deadline, check the exact model and product surface you plan to use, then check its displayed allowance and reset. Keep a fallback model available for routine work. If the plan reaches a limit, waiting or changing to an available model is the supported path; repeatedly switching between models does not create new allowance. Recheck plan pages before renewal or upgrade because tier names, bundled tools, and limits change. For API versus app billing, see subscription or API?.