Anthropic drops Opus 4.8 for Claude Code

Opus 4.8 cements Anthropic's lead in the AI coding space by aggressively targeting long-horizon workflows where prior models silently introduced bugs. Defaulting to high effort without heavily inflating token spend makes it a pragmatic upgrade for complex codebase tasks. The 69.2% SWE-bench Pro score represents a material jump over Opus 4.7, widening the gap on real-world coding evaluation. Being four times less likely to allow flaws in generated code means Claude Code can execute multi-step goals with less micromanagement. Bundled claude-api migration guidance and integration with /goal signals Anthropic is shifting focus from basic prompts to autonomous, goal-oriented agent loops.