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

Anthropic has released Claude Opus 4.8, an upgrade to its Opus class of AI models, available globally today. The release includes new features such as user-controlled effort levels, "dynamic workflows" in Claude Code, and a faster, cheaper mode. Anthropic reports improvements in coding, agentic tasks, reasoning, and honesty, alongside reduced rates of misaligned behaviour compared to the predecessor, Opus 4.7.

Pricing for standard usage remains unchanged from Opus 4.7, while fast mode is now three times cheaper. Standard pricing remains at $5 per million input tokens and $25 per million output tokens. Fast mode for Opus 4.8 operates at 2.5× the speed and costs $10 per million input tokens and $50 per million output tokens.

Claude Opus 4.8 is described as being around four times less likely than Opus 4.7 to allow flaws in code to pass unremarked. Early testers report the model is more likely to flag uncertainties and less likely to make unsupported claims, addressing issues where AI models jump to conclusions. The model defaults to "high effort," which Anthropic judges to be the best balance of quality and user experience.

Users can select "extra" (or "xhigh" in Claude Code) or "max" effort levels for difficult tasks and long-running asynchronous workflows. Rate limits in Claude Code have been increased to accommodate higher token usage associated with higher effort levels. Claude Code’s new "dynamic workflows" feature allows it to tackle very large-scale problems, including codebase-scale migrations across hundreds of thousands of lines of code.

Anthropic plans to release a new class of model with higher intelligence than Opus in the coming weeks, pending the development of stronger cyber safeguards. A small number of organisations are currently using Claude Mythos Preview for cybersecurity work as part of Project Glasswing. Additionally, Anthropic is opening a new office in Milan, its sixth in Europe.