Claude Rate Limits in 2026: Pro, Max, Claude Code & API

Claude does not have one universal rate limit. A Claude subscription, Claude Code signed in with that subscription, and the pay-as-you-go Claude API each use a different metering model. This guide explains which limit applies, where to see the reset, and what to check when Claude says you have reached a limit.

Facts checked August 9, 2026. Plan terms and model availability can change, so the linked Anthropic pages remain the source of truth for your account.

Claude Rate Limits: The Short Answer

• Claude Free, Pro, and Max: usage depends on the plan, model, conversation size, features, and effort level.
• Claude Code with a Claude subscription: usage is shared across Claude, Claude Code, and Claude Desktop.
• Claude Code with an API key: usage is pay-as-you-go and subject to Claude API rate and spend limits.
• Claude API: limits are enforced by requests per minute, input tokens per minute, output tokens per minute, and monthly spend caps.

Claude Free, Pro, and Max Usage Limits

Anthropic describes subscription usage as a conversation budget rather than a guaranteed number of messages. Long conversations, attachments, extended thinking, model choice, effort, and tool use can consume more capacity than short text prompts. That is why two users on the same plan can reach a limit after a different number of messages.

Pro provides more included usage than Free. Max is offered in 5x and 20x capacity tiers relative to Pro and can also have weekly limits, including an all-model limit and a Sonnet-specific limit. Claude shows the reset time attached to your account in Settings → Usage.

Avoid planning around unofficial fixed message counts. Anthropic can adjust capacity and does not promise that every prompt costs the same amount of usage. Check the official usage and length limits guide for the current behavior.

Does Claude Code Share the Same Limit?

It depends on how Claude Code is authenticated. If you sign in with a Pro or Max subscription, interactive Claude Code activity draws from the same included usage pool as Claude on the web, desktop, and mobile. A long repository task can therefore reduce the capacity available for a later Claude conversation.

If Claude Code is authenticated with an Anthropic API key, it uses pay-as-you-go API billing instead. The /status command helps subscription users inspect their allocation;/cost shows session token and dollar usage for API-key billing. AnANTHROPIC_API_KEY environment variable can cause Claude Code to use API billing even when you intended to use a subscription.

Anthropic documents the distinction in its Claude Code subscription guide.

Claude API Rate Limits: RPM, ITPM, and OTPM

The Claude API separates throughput limits from spending limits. The main Messages API limits are:

• RPM: requests per minute.
• ITPM: uncached input tokens per minute for most current models.
• OTPM: output tokens per minute as tokens are generated.

Claude Platform organizations are grouped into Start, Build, and Scale usage tiers, with Custom arrangements available for larger accounts. Actual limits can vary by model group, organization history, and workspace configuration. The Limits page in Claude Console is the authoritative view for your organization.

Anthropic uses a token-bucket system, so capacity replenishes continuously. A published per-minute limit may also be enforced over shorter intervals, which means a burst can trigger a 429 even when the minute-level average looks acceptable.

Claude API 429 Errors and Rate-Limit Headers

A 429 response means a request exceeded an active rate limit or acceleration limit. Read theretry-after header before retrying and use exponential backoff with jitter. Theanthropic-ratelimit-* headers expose the active limit, remaining capacity, and replenishment time for the most restrictive token constraint.

If traffic has increased sharply, ramp it gradually. If input tokens are the bottleneck, prompt caching can improve effective throughput because cached input does not count toward ITPM for most current Claude models. See Anthropic's API rate-limit reference for current model groups and headers.

Claude Console Spend Limits and Usage Monitoring

A spend limit is not the same as a rate limit. Spend limits cap monthly API cost; RPM, ITPM, and OTPM control throughput. Claude Console exposes both, along with Usage and Cost pages. Organizations can also use the Usage and Cost Admin API and the Rate Limits API for reporting and alerts instead of hardcoding values that may change.

For a subscription, optional usage credits can continue work beyond included limits at consumption-based rates. Those charges are separate from the subscription and can apply to both Claude conversations and Claude Code.

How to Use More Claude Without Hitting the Limit

• Start a fresh conversation when old context is no longer useful.
• Use /compact or /clear in Claude Code to control context growth.
• Select the least expensive capable model and use higher effort only when it adds value.
• Cache repeated API context and smooth bursty traffic.
• Check the reset shown in Claude before starting a long task.

See our focused guide to avoiding Claude Code rate limits for workflow-level tactics.

How to Track Claude Usage on a Mac

AIUsageBar keeps the Claude usage and reset information available to your account in the macOS menu bar, alongside other AI tools. Use Claude and Claude Console as the billing source of truth; use the menu bar view to notice pace and reset timing without repeatedly opening a dashboard.

Related: Claude usage tracker for Mac · Claude Code vs Claude Web limits.