Anthropic Rolls Back Claude Code Changes After Quality Decline; All Fixes Complete

Gate News message, April 24 — Anthropic has acknowledged a recent decline in Claude Code quality and confirmed that all related issues have been resolved through rollbacks and fixes. The problems stemmed from three product and prompt adjustments made between early and mid-April.

On March 4, the default reasoning effort was reduced from high to medium, which degraded performance until it was restored to high/xhigh on April 7. On March 26, a bug was introduced in the clear_thinking_20251015 logic designed to optimize caching, causing the system to lose reasoning history during extended idle periods and leading to memory loss and repetition; this was fixed in version 2.1.101 on April 10. Most recently, on April 16, a system prompt was added to compress output (limiting responses to ≤25/100 words), which resulted in approximately 3% negative impact on code quality; this change was rolled back on April 20.

Anthropric stated that the API reasoning layer was not affected by these issues. The company has also reset usage quotas for all subscribers.

Disclaimer: The information on this page may come from third parties and does not represent the views or opinions of Gate. The content displayed on this page is for reference only and does not constitute any financial, investment, or legal advice. Gate does not guarantee the accuracy or completeness of the information and shall not be liable for any losses arising from the use of this information. Virtual asset investments carry high risks and are subject to significant price volatility. You may lose all of your invested principal. Please fully understand the relevant risks and make prudent decisions based on your own financial situation and risk tolerance. For details, please refer to Disclaimer.

Related Articles

Nvidia to Invest Up to $2.1 Billion in Data Center Firm IREN

According to Bloomberg Terminal, Nvidia announced Thursday it will invest up to $2.1 billion in data center developer IREN Ltd. as part of a partnership to accelerate artificial intelligence infrastructure construction. IREN has granted Nvidia a five-year warrant to purchase as many as 30 million sh

GateNews1h ago

CZ Says YZi Labs Allocates 70% to Blockchain, 20% to AI at Consensus Miami 2026

According to ChainCatcher, at Consensus Miami 2026, Zhao Changpeng (CZ) stated that YZi Labs allocates 70% of funding to blockchain, 20% to AI, and 10% to biotechnology. CZ added that BNB should be positioned as a native currency for AI agents, and all blockchains need to be "AI ready" to support

GateNews7h ago

Public Acquires AI Investment Platform Treasury App to Expand Crypto Trading

According to ChainCatcher, Public announced the acquisition of AI investment service platform Treasury App to strengthen its AI-driven brokerage business. The acquisition amount was not disclosed. Public currently supports trading in stocks, bonds, and cryptocurrencies including Bitcoin, Ethereum, a

GateNews8h ago

Blitzy Completes $200M Funding Round Led by Northzone

According to ChainCatcher, Blitzy, an AI coding company co-founded by former Nvidia architect Sid Pardeshi, has completed a $200 million funding round led by Northzone. Battery Ventures, Jump Capital, and Morgan Creek Digital participated in the round. The platform can parse complex systems with

GateNews9h ago

EU Bans AI-Generated Deepfake Pornography on May 7

According to Xinhua News Agency, on May 7, EU Parliament members and member states reached consensus to ban artificial intelligence systems from generating deepfake pornographic content. The prohibition will be incorporated into amendments to the 2024 Artificial Intelligence Act. European Parliament

GateNews9h ago

Tether Releases QVAC MedPsy Medical AI Model, Achieves 62.62 Score on 17B-Parameter Version

According to Odaily, Tether AI Research Group released QVAC MedPsy, a medical AI model designed to run locally on smartphones and wearable devices without cloud dependency. The 1.7 billion-parameter version scored 62.62 on seven medical benchmarks, outperforming Google's MedGemma-1.5-4B by 11.42 poi

GateNews9h ago
Comment
0/400
No comments