Back to list
Anthropic Addresses Claude Code Quality Degradation Reports and Implements Fixes for Sonnet and Opus Models
Industry NewsAnthropicClaudeAI Engineering

Anthropic Addresses Claude Code Quality Degradation Reports and Implements Fixes for Sonnet and Opus Models

Anthropic has released a postmortem addressing recent user reports regarding the degradation of Claude's performance across specific tools, including Claude Code, the Claude Agent SDK, and Claude Cowork. The investigation identified three distinct technical issues occurring between March and April 2026: an intentional but poorly received reduction in reasoning effort to manage latency, a session-clearing bug that caused repetitive behavior and memory loss, and a system prompt change aimed at reducing verbosity that inadvertently harmed coding quality. While the API remained unaffected, these issues impacted Sonnet 4.6, Opus 4.6, and Opus 4.7. Anthropic has since reverted the problematic changes and fixed the bugs as of April 20 (v2.1.116), emphasizing their commitment to maintaining model intelligence over speed.

Hacker News
Share

Key Takeaways

  • Three Distinct Issues Identified: The perceived degradation was caused by a change in reasoning effort, a session-clearing bug, and a system prompt instruction to reduce verbosity.
  • Specific Tools Affected: Issues were limited to Claude Code, the Claude Agent SDK, and Claude Cowork; the core API and inference layer were not impacted.
  • Models Impacted: The performance dips affected Sonnet 4.6, Opus 4.6, and Opus 4.7 across different timeframes.
  • Full Resolution: All identified issues were resolved as of April 20 with the release of version 2.1.116.

In-Depth Analysis

Reasoning Effort and Latency Trade-offs

On March 4, Anthropic attempted to address UI latency issues where the interface appeared frozen by changing the default reasoning effort from "high" to "medium." While this was intended to improve the user experience by reducing wait times, it resulted in a noticeable drop in intelligence for Sonnet 4.6 and Opus 4.6. Following user feedback indicating a preference for higher intelligence over speed, Anthropic reverted this change on April 7. The company acknowledged that prioritizing lower latency at the expense of reasoning quality was the "wrong tradeoff."

Technical Bugs and Prompting Side Effects

Two additional technical factors contributed to the degradation. On March 26, a feature designed to clear old thinking from idle sessions to improve resumption speed introduced a bug. This bug caused the system to clear thinking every turn, making the models appear forgetful and repetitive. Furthermore, an April 16 update to the system prompt intended to reduce verbosity negatively impacted coding quality when combined with other prompt adjustments. This specific issue affected the latest models, including Opus 4.7. Both the bug and the prompt changes were corrected and reverted by April 20.

Investigation Challenges and Aggregate Effects

Anthropic noted that because these three changes occurred on different schedules and affected different segments of traffic, the resulting feedback appeared as broad and inconsistent degradation. The investigation began in early March but was complicated by the difficulty of distinguishing these specific technical failures from the normal variation in user feedback. The company has reaffirmed that they never intentionally degrade models and are implementing changes to prevent similar regressions in the future.

Industry Impact

This incident highlights the delicate balance AI providers must maintain between model "intelligence" (reasoning effort) and operational performance (latency). For the AI industry, it serves as a case study in how minor optimizations—such as reducing verbosity or clearing session cache—can have significant, unintended consequences on the quality of complex tasks like coding. Anthropic's transparent postmortem underscores the importance of user feedback loops in identifying non-obvious regressions that automated testing might miss, particularly when those regressions are tied to UI-specific implementations rather than the underlying API.

Frequently Asked Questions

Question: Was the Claude API affected by these quality issues?

No. Anthropic confirmed that the API and inference layer remained unaffected throughout this period; the issues were isolated to Claude Code, the Claude Agent SDK, and Claude Cowork.

Question: Which Claude models were impacted by the performance degradation?

The issues affected Sonnet 4.6, Opus 4.6, and Opus 4.7, depending on the specific technical change and the timeframe.

Question: How has Anthropic resolved these issues?

As of April 20 (v2.1.116), Anthropic has reverted the reasoning effort to "high," fixed the session-clearing bug, and removed the system prompt instructions that were harming coding quality.

Read original article

Related News

Industry News

Perplexity Deploys GPT-6 Astra Across Critical End-to-End Workflows and Production Systems

According to an update published by OpenAI, Perplexity is leveraging the advanced capabilities of GPT-6 Astra across core end-to-end organizational and technical systems. The implementation spans multiple operational domains, with Perplexity using Astra to draft internal and external communications, modify and update software codebases, and maintain active monitoring over production environments. A notable shift in operational management highlighted in the report is that teams at Perplexity now require substantially fewer check-ins compared to their workflows with earlier artificial intelligence models. This adoption marks a significant milestone in software engineering and system oversight, demonstrating how higher-reliability model architectures enable organizations to delegate mission-critical maintenance and development tasks with less human intervention while sustaining production stability.

Waymo Robotaxi Pulls Over and Alerts San Francisco Police to Detain Armed Juvenile Riders
Industry News

Waymo Robotaxi Pulls Over and Alerts San Francisco Police to Detain Armed Juvenile Riders

In San Francisco, an autonomous Waymo vehicle pulled over and contacted law enforcement after detecting unauthorized activity involving two juvenile passengers. The incident resulted in the arrest of the two minors, who were found in possession of a loaded AR-style ghost gun inside the robotaxi cabin. Both suspects were taken into custody and transported to a juvenile hall. While initial police documentation did not explicitly identify who was operating or managing the vehicle during the ride, the event underscores the operational capabilities of autonomous fleets to monitor cabin security, respond to severe policy violations, and autonomously coordinate with local law enforcement to maintain public safety.

Hyundai Postpones In-House AI Driver-Assist System Launch to 2029 While Partnering with Nvidia
Industry News

Hyundai Postpones In-House AI Driver-Assist System Launch to 2029 While Partnering with Nvidia

Hyundai Motor Group has delayed the debut of its proprietary artificial intelligence driver-assistance software to late 2029, pushing back its original schedule by roughly two years. To maintain commercial competitiveness in the interim, the South Korean automaker is deepening its technical collaboration with Nvidia to roll out advanced Level 2+ and Level 2++ driver-assistance platforms starting in 2028. Hyundai's initial lineup of Nvidia-based vehicles will bypass expensive lidar sensors, relying instead on an integrated suite of optical cameras, radar units, and ultrasonic sensors. While lidar remains under active consideration for higher-tier Level 3 automated driving systems, Hyundai aims to utilize real-world driving data collected from its 2028 commercial fleet to train, validate, and mature its proprietary in-house autonomous software platform ahead of its rescheduled 2029 release.