Anthropic released Claude Opus 4.8, improving on Opus 4.7 across coding, agentic skills, reasoning, and knowledge work, and reporting it was roughly four times less likely to overlook a flaw in code. It also cited better "honesty," flagging uncertainty rather than making unsupported claims. It scored 84% on the Online-Mind2Web web-navigation benchmark and became the first model to exceed 10% on a legal agent benchmark's all-pass standard. Claude Code gained dynamic workflows for large-scale problems, and claude.ai gained effort control for dialing computational intensity. Pricing matched Opus 4.7 at $5 and $25 per million tokens, with a 2.5x fast mode at $10 and $50 — a third the price of fast mode on previous models. Anthropic announced its Series H the same day.