Anthropic Releases Claude Opus 4.8 for Coding and Long-Horizon Work

Anthropic introduced Claude Opus 4.8, describing stronger performance across coding, agentic tasks and professional work, with an emphasis on consistency for long-running assignments. The release came as vendors competed on how well models could carry context and tools across a sequence of actions.

For teams evaluating agents, long-horizon consistency is a more useful claim than a single-turn demo. It affects whether an assistant can complete a multi-step change, recover from partial failure and deliver output that a reviewer can validate without rebuilding the work.