Introducing Claude 4 (www.anthropic.com)

Anthropic has unveiled Claude 4, featuring two new models: Claude Opus 4 and Claude Sonnet 4, setting new benchmarks in coding, advanced reasoning, and AI agent capabilities. Claude Opus 4 emerges as the world’s premier coding model, excelling in sustained, complex workflows and delivering top scores on industry benchmarks like SWE-bench (72.5%) and Terminal-bench (43.2%). Its ability to maintain performance over multi-hour tasks marks a significant leap in AI’s capacity for long-running, intricate projects. Claude Sonnet 4, an upgrade over Sonnet 3.7, balances power and efficiency with improved instruction-following and code quality, scoring 72.7% on SWE-bench and gaining adoption in tools like GitHub Copilot for agentic coding scenarios. Key technical innovations include “extended thinking” with parallel tool use, enabling the models to interleave reasoning and tool interactions such as web search for richer responses. Both models show 65% less reliance on task shortcuts and exhibit advanced memory through local file access, allowing long-term context management and improved task continuity. The newly generally available Claude Code integrates deeply with developer environments via VS Code and JetBrains, supports GitHub Actions for background tasks, and offers an extensible SDK to build custom AI-powered workflows. Accessible via Anthropic’s API, Amazon Bedrock, and Google Cloud Vertex AI, these hybrid models support fast responses and longer reasoning chains at existing pricing tiers. Claude 4 represents a major step toward reliable AI collaborators capable of handling complex, multi-step coding and reasoning challenges at scale.