Anthropic unveils Claude Opus 4 and Sonnet 4, new AI models for advanced coding and reasoning

Anthropic has launched its next-generation AI models—Claude Opus 4 and Claude Sonnet 4—touting major advances in coding, reasoning, and long-running agent workflows. With these upgrades, the Claude lineup aims to offer better models for developer tools and autonomous AI agents.

Claude Opus 4, now positioned as the most powerful model in Anthropic’s arsenal, is being internally hailed as the “world’s best coding model.” It leads key benchmarks like SWE-bench (72.5%) and Terminal-bench (43.2%), outperforming rivals in tasks that require sustained effort and nuanced code comprehension. According to early adopters like Cursor, Replit, and Block, Opus 4 enables advanced editing, debugging, and multi-file code modifications with high reliability. Rakuten and Cognition further validated its capabilities with prolonged, uninterrupted coding tasks spanning hours.

Meanwhile, Claude Sonnet 4 delivers a significant performance jump over its predecessor, Sonnet 3.7, achieving a 72.7% score on SWE-bench. Though not as powerful as Opus 4, Sonnet 4 focuses on balancing efficiency and performance, making it ideal for general use cases. GitHub has already tapped Sonnet 4 to power a new version of GitHub Copilot, citing improvements in instruction-following and problem-solving. Sourcegraph and other tech firms noted its superior code quality, deeper problem understanding, and enhanced agent reliability, as per Anthropic.

Both models boast enhanced memory features, particularly Opus 4, which can now create “memory files” for persistent context—improving its ability to handle complex, long-term agent tasks.