Claude Opus 4.8 launched with new agentic features, honesty upgrade, and cheaper speed

Anthropic has unveiled a new version of its most powerful model. The Claude Opus 4.8 builds on version 4.7 and it focuses heavily on sharper reasoning, long-running autonomous tasks, stronger honesty, and a set of more features. The best part? Pricing remains unchanged despite various improvements. Yesterday, we reported that Anthropic is working on adding support for multiple languages to Claude’s Voice Mode. 

Claude Opus 4.8 focuses on reliability and agentic workflows

The headline improvement in Claude Opus 4.8 is judgement and transparency. Anthropic claims that the latest version is more likely to flag any uncertainties and less likely to make unsupported claims compared to earlier versions. This means fewer hallucinated updates and more reliable outputs. 

For developers using Claude in agentic workflows, the upgrade is tangible. Opus 4.8 can deliver stronger results across coding, reasoning, financial analysis, and professional knowledge work. Anthropic is also introducing dynamic workflows in Claude Opus 4.8.The feature allows Claude Code to plan and run hundreds of parallel sub-agents in a single section for handling large and complex tasks. The company says that this should help developers automate long-running engineering and productive workloads more effectively.

Users on claude.ai also get a new effort control to choose how deeply Claude thinks per response, trading speed for quality depending on the task.

Claude Opus 4.8 is available today everywhere via the Claude API. It costs $5 per million input tokens and $25 per million output tokens. Anthropic also confirmed that Mythos class model is expected to reach general availability in the coming weeks.

Source: Claude