Anthropic Launches Claude Opus 4.8 With New Agentic Features and Improvements

Anthropic's most powerful model available to the public is getting an upgrade with the release of Claude Opus 4.8, which has been built on the previous 4.7 version and focuses on one key feature.

The latest update from Anthropic focuses on upgrading its agentic features on the model, particularly by offering improvements in carrying out agentic tasks for users, as well as several new additions.

Anthropic Debuts Claude Opus 4.8 With Agentic Upgrades

Anthropic has officially announced the debut of its new and improved model, the Claude Opus 4.8, which the company says is a new era for its agentic AI-focused model. According to Anthropic, Claude Opus 4.8 is now "a more effective collaborator" thanks to its latest improvements across its benchmarks and specific features.

The company stresses that this upgrade will deliver improved collaboration features as it is more reliable and sharper "in its judgment when it's performing agentic tasks."

Anthropic added that it also upgraded a trait that Opus is made for, which is honesty. The upgrade ensures that all of its output is based on facts and information that can be backed up.

Anthropic said that Claude Opus 4.8 will not jump to conclusions, and it will flag uncertainties in its work, making it four times less likely to have flaws in its code compared to its predecessor, Claude Opus 4.7. Moreover, Claude Opus 4.8 is "prosocial" as it better supports user autonomy and "acting in the user's best interest."

Other New Features from Anthropic Alongside Opus 4.8

According to 9to5Google, Anthropic also has various other improvements to this new model apart from its massive focus on upgrading its agentic AI capabilities.

Anthropic explained that there are three additional features that are part of the 4.8 version to help users get the most out of their access.

First, there is Dynamic Workflows, which is currently in research preview, and it brings the capabilities for the model to take larger tasks in the Claude Code feature.

Under the feature, Claude can plan the work and proceed to running "hundreds of parallel subagents in a single session," then verify the output all before reporting back to users. Dynamic Workflows is available in Claude Code for Enterprise, Team, and Max plans.

Next, users may now tweak "Effort control" in the Claude website and Cowork, with this feature helping users to choose the specific effort level for a response, with higher effort translating to deeper thinking and better responses. Effort control is available on all plans, according to Anthropic.

Lastly, developers can now update Claude's instructions mid-task as Anthropic said that it will no longer break the prompt cache for a system entry with the Messages API.