Anthropic drops Claude Opus 4.8

Opus 4.8 establishes a terrifying new baseline for autonomous engineering and computer use capabilities.

• –69.2% on SWE-Bench Pro suggests it can resolve the vast majority of real-world software issues without human intervention
• –83.4% on agentic computer use indicates a massive leap in its ability to directly drive desktop applications
• –Beating GPT-5.5 across the board solidifies Anthropic's lead in the frontier model race
• –Scoring 57.9% on Humanity's Last Exam with tools highlights advanced reasoning on complex edge cases