An update on recent Claude Code quality reports
anthropic.com
April 23, 2026
8 min read
🔥🔥🔥🔥🔥
68/100
Summary
Recent reports indicated a decline in response quality for some users of Claude Code, Claude Agent SDK, and Claude Cowork due to three specific changes. These issues have been resolved as of April 20 with the release of version 2.1.116, while the API remained unaffected.
Key Takeaways
- Claude's response quality issues were traced to three separate changes affecting Claude Code, the Claude Agent SDK, and Claude Cowork, but the API remained unaffected.
- As of April 20, all identified issues have been resolved with the release of version 2.1.116.
- User feedback led to a change in default reasoning effort settings, reverting to high effort for Opus 4.7 after reports of decreased intelligence with medium effort.
- Usage limits for all subscribers were reset on April 23 to improve user experience following the quality reports.
Community Sentiment
Positives
- Running GPT-5.4 at extra high effort has resulted in minimal mistakes, showcasing its improved reasoning capabilities and reliability for critical data integrity tasks.
- Claude's recent updates aim to reduce latency, which is crucial for maintaining user engagement and responsiveness in interactive applications.
- The model's ability to preemptively follow critical instructions demonstrates significant advancements in understanding user intent and context.
Concerns
- Claude's tendency to respond to its own internal prompts raises concerns about the model's reliability and could lead to confusion in user interactions.
- The recent change to purge older context after an hour compromises the model's ability to maintain continuity in longer sessions, which is essential for complex problem-solving.
- Users are experiencing a perceived drop in quality, suggesting that the non-deterministic nature of the model's output may lead to inconsistent performance.
