Anthropic launches Claude Opus 4.8. The model scores better on benchmarks and offers new features such as dynamic workflows and more efficient tool use at the same price.
Claude Opus 4.8 is now available to all users and builds upon its predecessor, Opus 4.7. The model performs better in areas such as coding, reasoning, and agentic tasks. Additionally, Anthropic is introducing new features that improve collaboration with the model.
According to testers, Opus 4.8 stands out due to its improved judgment and reliability. The model completes tasks faster and more accurately, particularly in complex, multi-service explorations and legal work.
New features and improved performance
Claude Opus 4.8 introduces dynamic workflows in Claude Code. This allows the model to break down large tasks into hundreds of parallel sub-agents that work independently and verify results. This feature is available in the Enterprise, Team, and Max plans of Claude Code.
Additionally, users can now adjust the effort Claude puts into a task via claude.ai and Cowork. With a higher effort setting, the model thinks more deeply for better answers, while a lower setting provides faster responses. Opus 4.8 defaults to high effort, providing a good balance between quality and user experience.
Availability and future plans
Claude Opus 4.8 is available today at the same price as Opus 4.7: five dollars per million input tokens and 25 dollars per million output tokens. The fast mode is now three times cheaper than with previous models.
Anthropic is also working on a new class of models with even higher intelligence than Opus, under the name Project Glasswing. A limited number of organizations are currently testing Claude Mythos Preview for cybersecurity applications. The intention is to make these models available to all customers in the coming weeks.