What happened
Anthropic reset Claude’s five-hour and weekly rate limits for all users on July 16. It had previously reset the same limits on July 10. Neither announcement defined how long the extra capacity would last.
The resets happened while OpenAI launched GPT-5.6 and restored Codex allowances. That timing made capacity part of the competition between coding agents. It does not prove either company reacted to the other.
Fable 5 followed a similar sequence. Anthropic first included it temporarily, then extended access twice. It has now published a longer-term arrangement: Max and Team Premium plans include Fable 5 at 50% of limits from July 20. Pro and Team Standard users continue through usage credits.
Why it matters
Claude’s usage limits are part of the product, not a footnote. Anthropic says Claude.ai, Claude Code, and Claude Desktop draw from the same usage budget. That means a builder’s decision to test a model in chat can affect how much capacity remains for coding later.
The distinction is between recurring and promotional capacity. A plan entitlement describes what a subscriber should expect after the promotion ends. A reset gives users more room now, but it does not change the normal allowance unless the vendor says so.
The practical competition between coding agents is not only about model quality. It is also about whether users can try a new model, recover from a limit, and keep a workflow running when a rival launches something important.
That matters especially for builders without a team budget. A developer choosing between Claude Code and Codex may care less about a benchmark gap than about whether one tool stops in the middle of a long debugging session.
Who should care
Anyone using a paid Claude plan to build software should pay attention to the difference between a normal rolling reset and a vendor-wide reset. The former is part of the published usage system. The latter changes the amount of work users can complete immediately, even if the plan’s price and stated limits do not change.
Teams should also separate model evaluation from quota timing. A fresh limit can make a tool feel better simply because it lets you run more tasks in a short period. That is useful, but it is not the same as proving that the model produces better code.
What builders should do next
Use the reset for a bounded task that normally approaches your limit. Define the expected result first. Then record completion quality, corrections, elapsed time, and allowance consumed.
Check Claude’s usage page before and after the test. Anthropic says usage across Claude.ai, Claude Code, and Claude Desktop counts toward the same limit, so a chat experiment can reduce the allowance available for coding.
Evaluate a subscription using its recurring allowance, not a promotional reset. Temporary capacity is useful, but only the normal limit tells you what the plan can sustain.
End of article