Anthropic releases Claude Opus 4.8, emphasizing the model's improved ability to acknowledge uncertainties and avoid unsupported claims

Anthropic is releasing Claude Opus 4.8 on Thursday. According to the company, the model is more likely to flag uncertainties about its work and less likely to make unsupported claims compared to its predecessor.

In Anthropic's evaluations, Opus 4.8 is around 4x less likely than its predecessor to allow flaws in code it has written to pass unremarked. Users can also direct the amount of effort Claude puts into a task, with higher-effort responses consuming more tokens.

Anthropic is launching a feature called 'dynamic workflows' in research preview. With this feature, Claude can plan work and run hundreds of parallel subagents (independent AI agents that execute tasks) in a single session, then verify outputs before reporting back to the user.