Claude hands paid users a spare limit reset until Oct 22

Anthropic has given every Pro, Max and Team subscriber one extra usage reset. It sits in Settings → Usage and can be applied any time until October 22, the company’s ClaudeDevs account posted on Twitter on September 22. Claude Code limits changed in two other ways the same day, and each change touches a different limit.

The five-hour session limit rose 20%. The second change comes from the model. Opus 5.5 is now the default on paid plans, and because it is priced below Opus 5, Anthropic says both the five-hour and the weekly limits go 25% further on it. Neither figure is a token count. Anthropic publishes only the ratios between plans, and how far an allowance stretches depends on message length, attached files, conversation length, the model and demand.

The weekly limit has moved a lot this month. A promotion that added 50% to Claude Code’s weekly limits ran from May 13 to September 13. From September 14 the weekly limit sits 25% above the old baseline, which against the promotion is a 17% cut, and eight days later the 25% that Opus 5.5 adds depends on running the cheaper model.

Testing Catalog spotted the Reset button on Claude web and Desktop as well, with the label “explore Opus 5.5!”. On Pro and Max, Claude Code has no quota of its own. It shares one pool with Claude.ai chat and Cowork, under the same five-hour and weekly limits, so a reset applied from the web app likely frees up Claude Code too.

On the Claude Platform, Opus 5.5 costs $4 per million input tokens and $20 per million output tokens. Opus 5 cost $5 and $25, so both list prices drop 20%. Cache reads fall 60%, from $0.50 to $0.20 per million. Anthropic says Opus 5.5 performs at the level of Fable 5.1 and runs about 30% faster and about 40% cheaper per task than Opus 5. The gap between the 20% list cut and the 40% per-task figure probably comes mostly from cheaper cache reads, since long agent sessions reread the same context over and over.

Thinking is always on for Opus 5.5. The default effort is medium, and you can change it with the effort parameter. For long-running agents, compaction summarizes older turns and keeps recent ones intact. If you need more speed, fast mode runs up to 2.5x faster at $8/$40 per million tokens.

Anthropic also says Opus 5.5 writes more clearly, uses less jargon and follows the writing rules you give it, which it says makes long Claude Code sessions easier to follow.

One change can interrupt a session. Opus 5.5 is the first Opus with Fable 5.1-class safeguards on cyber, bio and frontier LLM development. A request the safeguards flag falls back to another model, and Anthropic says it is working to reduce incorrect flags.

To move an API project over, run /claude-api migrate in Claude Code, which switches it to claude-opus-5-5.