AI is changing the world. Don't stay behind. Clear summaries, community insight, delivered without the noise. Subscribe to never miss a beat.

Anthropic silently downgraded cache TTL from 1h → 5M on March 6th

github.com

April 12, 2026

8 min read

🔥🔥🔥🔥🔥

68/100

Summary

Cache TTL for Claude Code regressed from 1 hour to 5 minutes around early March 2026. This change resulted in significant quota and cost inflation.

Key Takeaways

The default cache TTL for Claude Code was changed from 1 hour to 5 minutes around early March 2026, leading to increased costs.
This change resulted in a 20–32% increase in cache creation costs and a spike in quota consumption for users.
Analysis of session data from January to April 2026 confirmed the regression in cache TTL behavior across two independent machines.
The transition to a dominant 5-minute TTL occurred between March 6 and March 8, 2026, after a consistent period of 1-hour TTL usage.

Community Sentiment

Negative

Positives

Codex has shown remarkable performance, with users expressing awe at its capabilities compared to Claude, indicating a strong preference for its current functionality.
Switching from Claude to Codex has proven beneficial for several users, who find it delivers the same quality of output with a more manageable subscription model.

Concerns

There is a growing sentiment of distrust among users regarding Claude's performance, as many feel they are not receiving the product they initially paid for, leading to negative perceptions.
Users have reported a noticeable decline in Claude's reasoning capabilities, with some experiencing illogical outputs and a lack of clarity in its responses.
The recent changes, including the downgrading of cache TTL, have raised concerns about potential undisclosed limitations and performance regressions in Claude.