Anthropic Releases Claude Opus 4.8 with Enhanced Coding and Agentic Capabilities
Anthropic has launched Claude Opus 4.8, an upgraded version of its Opus model, featuring significant performance improvements in coding, agentic tasks, and long-running work. This release includes new features like effort control for users and dynamic workflows in Claude Code for large-scale problem-solving. The model is available today and demonstrates enhanced judgment, reliability, and efficiency across various benchmarks, making it a more effective collaborator for professional and engineering workloads.
- →Claude Opus 4.8 introduces stronger performance across key areas
- →New 'dynamic workflows' feature for Claude Code enables large-scale problem-solving
- →Effort control introduced for users on claude.ai and Cowork
- →Faster and more cost-effective 'fast mode' for Opus 4.8
- →Improved honesty and reduced unremarked flaws in Opus 4.8
Features (3) ›
- Claude Opus 4.8 introduces stronger performance across key areas
The new Claude Opus 4.8 model offers improved performance in coding, agentic tasks, and professional work, along with enhanced consistency for long-running tasks, building on the capabilities of its predecessor, Opus 4.7.
- New 'dynamic workflows' feature for Claude Code enables large-scale problem-solving
Claude Code now features 'dynamic workflows,' a research preview capability that allows the model to tackle very large-scale problems by planning and running hundreds of parallel subagents within a single session, verified before reporting outputs.
- Effort control introduced for users on claude.ai and Cowork
Users can now control the amount of effort Claude expends on a task, with higher settings leading to deeper thinking and better responses, while lower settings offer faster responses and slower rate limit consumption.
Enhancements (3) ›
- Faster and more cost-effective 'fast mode' for Opus 4.8
The fast mode for Claude Opus 4.8 operates at 2.5 times the speed and is now three times cheaper than for previous models, offering a more economical option for rapid task completion.
- Improved honesty and reduced unremarked flaws in Opus 4.8
Claude Opus 4.8 demonstrates improved honesty, being four times less likely to allow flaws in its written code to pass unremarked and more likely to flag uncertainties about its work.
- Messages API now accepts system entries within the messages array
Developers can now update Claude’s instructions mid-task using system entries inside the messages array in the Messages API, allowing for real-time updates to permissions, token budgets, or environment context without breaking the prompt cache.
https://www.anthropic.com/news/claude-opus-4-8