Anthropic releases Claude Opus 4.8 with focus on AI honesty and dynamic workflows

Anthropic has released Claude Opus 4.8, an update to its artificial intelligence model that prioritises "honesty" as a core upgrade. The company states that the model is trained to avoid making unsupported claims and is more likely to flag uncertainties in its work, addressing a common issue where AI models jump to conclusions despite thin evidence.

According to internal evaluations cited by Anthropic, Opus 4.8 is approximately four times less likely than its predecessor to overlook flaws in generated code. The lab notes that early testers have found the new model is less prone to presenting work as progressing when evidence is lacking, aiming to improve reliability for users relying on the system for complex tasks.

The release also introduces adjustable effort levels, allowing users to direct the amount of computational effort Claude applies to a specific task. Higher-effort responses consume more tokens, while lower-effort options are available to help users conserve rate limits, providing greater control over resource allocation during development cycles.

A key feature of the update is the research preview of "dynamic workflows," which enables the AI to plan work and execute hundreds of parallel subagents within a single session. This capability allows Claude to manage larger, more complex tasks by running subagents for longer durations and verifying outputs before reporting back to the user.

Anthropic highlights that Claude Code with Opus 4.8 can now perform codebase-scale migrations across hundreds of thousands of lines of code. The company continues to accelerate its model release cycle, with this update following previous iterations designed to enhance the utility and safety of its artificial intelligence tools.