Claude Sonnet 4 degraded performance quality

Incident Report for Anthropic

Resolved

From 08:45 UTC on July 8th to 02:00 UTC on July 10th, Claude Sonnet 4 experienced a degradation in quality for some requests. Users, especially tool use and Claude Code users, would have seen lower intelligence responses and malformed tool calls.

This was caused by a rollout of our inference stack, which we have since rolled back. While we often make changes intended to improve the efficiency and throughput of our models, our intention is always to retain the same model response quality.
Posted Jul 10, 2025 - 02:00 UTC