Claude Opus
Tracing the Opus flagship line: from 4.5's coding and agent focus, through 4.6's million-token context, to 4.8's long-horizon agents.
Anthropic's flagship line moves in small steps. Opus 4.5 led with coding and agents and handed users an effort control; 4.6 became the first Opus with a million-token context; 4.8 shifted weight onto long-horizon agents and stressed not pretending to finish when evidence is thin. The surrounding changes say more than scores: parallel subagents in Claude Code, updatable system entries, a repriced fast mode.
