Anthropic released Claude Opus 4.8 on Thursday, its most advanced publicly available model, just 41 days after Opus 4.7—an unusually fast upgrade cycle amid growing competitive pressure from OpenAI and Google.

The new model maintains the same pricing as its predecessor at $5 per million input tokens and $25 per million output tokens, while delivering meaningful improvements in agentic tasks, coding, and notably, honesty. Opus 4.8 is around four times less likely than Opus 4.7 to let flaws in generated code pass unremarked, and early testers praised its tendency to proactively flag uncertainties rather than making unsupported claims.

Alongside the model, Anthropic launched Dynamic Workflows in research preview, enabling Claude Code to plan and run hundreds of parallel subagents in a single session—capable of handling codebase-scale migrations across hundreds of thousands of lines of code. The update also introduces effort controls on claude.ai, letting users balance response quality against speed and rate limits.

Anthropic also hinted that its most advanced model, Claude Mythos Preview, currently restricted to cybersecurity research due to safety concerns, could reach general availability within weeks as necessary safeguards are finalized.