Anthropic launches new Claude 4 models with advanced coding and reasoning capabilities

Anthropic has launched two new models, Claude Opus 4 and Sonnet 4, expanding capabilities for coding, advanced reasoning, and AI agent workflows. Claude Opus 4 is positioned as Anthropic’s most capable model, excelling at long-running, focused tasks over several hours. This model dramatically outperforms previous Sonnet versions, particularly in scenarios that demand thousands of steps or sustained effort.

While Claude Opus 4 expands possibilities for research, writing, and scientific discovery, Sonnet 4 brings notable upgrades over Sonnet 3.7 for everyday use. It features enhanced coding abilities, improved efficiency, and greater implementation control. Although Sonnet 4 does not match Opus 4 in all domains, it serves as an effective, efficient replacement for existing Sonnet 3.7 users. Following these upgrades, both models benefit from improved extended thinking, parallel tool execution, and memory management, with reductions in behaviors that exploit shortcuts or loopholes.

Additionally, developers now gain wider adoption potential with Claude Code, which moves to general availability. Claude Code extends Claude’s abilities to workflows in the terminal, integrated development environments, and background jobs via the Claude Code software development kit. To complement this, new beta extensions offer direct Claude Code integration in Visual Studio Code and JetBrains IDEs.