Anthropic Launches Claude Opus 4.8 With Gains in Coding and Honesty

Anthropic launched Claude Opus 4.8, claiming better agentic coding, reasoning, computer use, and financial analysis, with fewer unsupported claims.

The model posted benchmark gains against several rivals, though GPT-5.5 still leads on terminal coding.

Anthropic also added dynamic workflows in Claude Code, an effort-control setting in Claude.ai and Cowork, and a Messages API update for system instructions inside message arrays.

Opus 4.8 is available now at unchanged standard pricing, with a faster and cheaper fast mode.