Claude Code Pro Max 5x plan quota exhausted in just 1.5 hours even with moderate usage

On the Pro Max 5x (1M context) plan, even ordinary levels of Q&A and development work trigger a token limit exceeded in just 1.5 hours.

The cause is pointed to a bug where cache_read tokens are counted at the full rate (1.0x). Because tokens that should be saved through caching are counted at a much higher value than they actually are, the caching benefit effectively disappears in practice and usage is consumed rapidly.

Key points

• The consumption rate is abnormally fast even on the 1M context plan
• The limit is reached with only moderate levels of Q&A and development work
• The cache_read calculation error is cited as the cause
• The core of the problem is a user experience where the caching benefit seems to be nullified