Claude Opus 4.8
anthropic.com
May 28, 2026
8 min read
🔥🔥🔥🔥🔥
81/100
Summary
Claude Opus 4.8 introduces enhancements over version 4.7, improving benchmark performance and collaboration capabilities. New features include user-controlled effort levels, a "dynamic workflows" capability in Claude Code for large-scale problem-solving, and a fast mode for quicker task execution.
Key Takeaways
- Claude Opus 4.8 features improvements in reliability and judgment, making it more effective for agentic tasks compared to previous models.
- The model achieves the highest score on the Legal Agent Benchmark, breaking 10% overall on the all-pass standard, indicating significant accuracy for legal work.
- Opus 4.8 operates at 2.5Ă— the speed of previous models and is now three times cheaper in fast mode, enhancing cost efficiency for users.
- Early testers report that Claude Opus 4.8 delivers higher quality analysis and proactively flags issues, resulting in a better signal-to-noise ratio in outputs.
Coverage & discussions
Hacker News · anthropic.com
Community discussions
Hacker News discussion ↗
Historical count: 5 · 5 saved in sample · Historical sample
Analysis ready
Community Sentiment
Positives
- Opus 4.8 introduces notable improvements in honesty, which could enhance user trust and transparency in AI interactions.
- The potential for smaller models to outperform larger ones in coding tasks suggests a shift towards more efficient AI development, making advanced capabilities more accessible.
- Improvements in the context window expansion from previous versions have contributed to productivity gains, indicating a focus on enhancing user experience.
- The ongoing development of Project Glasswing and the promise of higher intelligence models like Claude Mythos could lead to significant advancements in cybersecurity applications.
Concerns
- Users report that the incremental updates from 4.5 to 4.8 have not resulted in noticeable improvements, raising concerns about the pace of innovation.
- The regression in reliability observed in version 4.7 has led some users to revert to earlier models, highlighting potential quality control issues.
- The exclusion of several key benchmarks in the latest release raises questions about the transparency and validity of the reported performance metrics.
- Concerns about access to advanced models like Claude Mythos suggest a risk of creating a divide in AI capabilities between large organizations and average users.
