Anthropic today launched the latest version of its flagship Opus model: Opus 4.5. The company calls it its most intelligent model yet and notes that it is especially strong in solving coding tasks, taking the crown from OpenAI’s GPT-5.1-Codex-Max and Google’s week-old Gemini 3 model with an SWE-Bench Verified accuracy score of 80.9%. The company is also making Opus 4.5 significantly more affordable to use, with API pricing of $5 per million input tokens and $25 per million output tokens, down fr...