Anthropic Launches Claude Opus 4.8 with Dynamic Workflows

Anthropic Launches Claude Opus 4.8 and Its Innovations

On May 28, 2026, Anthropic announced the release of Claude Opus 4.8 via a blog post. This new version arrives just six weeks after its predecessor, marking a faster-than-usual iteration pace. Claude Opus 4.8 is already available on the API and Claude Code, with an "unchanged price" for regular users, according to the company. This update brings several features, including dynamic workflows.

A More Reliable and Honest AI Model

Built on the foundations of Claude Opus 4.7, the 4.8 model improves performance across various benchmarks that assess skills such as reasoning and practical task execution. Early testers have noted a significant improvement in the reliability and accuracy of Claude Opus 4.8's judgments. Anthropic highlights the enhancement of the model's "honesty." A recurring issue with AI models is their tendency to draw hasty conclusions without sufficient evidence. However, Claude Opus 4.8 seems to avoid this pitfall, being more inclined to report uncertainties in its work and less likely to make unsupported claims.

Tom Pritchard, an engineer at Shopify, previewed Claude Opus 4.8 and confirms these improvements. According to him, in Claude Code, the model asks the right questions, detects its own mistakes, and pushes back when plans lack solidity. It also consolidates its understanding of complex multi-service explorations before making significant changes.

Advanced Features for Better Task Management

Alongside this update, Anthropic is introducing a new feature in Claude Code called Dynamic workflows. Currently in research preview, this feature allows Claude to plan tasks, launch hundreds of sub-agents in parallel to process them, and then verify the results before submitting them to the user. For example, Claude Code with Opus 4.8 can now manage migrations at the scale of an entire codebase, involving hundreds of thousands of lines, from launch to merge, using the existing test suite as a validation criterion.

The Dynamic workflows are available in research preview for subscribers to the Enterprise, Team, and Max plans. Additionally, a feature named Effort control has been rolled out to all users on Claude and in Cowork. This feature allows users to adjust "the level of effort that Claude puts into a response," according to Anthropic. With high effort, Claude thinks more frequently and deeply, while with reduced effort, it responds faster and consumes rate limits more slowly. This option is particularly useful for managing token consumption, as users seek to avoid session throttling. To adjust the level of thought, simply click on the model's name and then on Effort.