Anthropic Releases Claude Opus 4 and Sonnet 4 with Enhanced Coding and Controversial Safety Features
Opus 4 excels at coding and can run autonomously for seven hours, while Sonnet 4 powers GitHub Copilot.
Anthropic has launched the Claude 4 model family, headlined by the powerful Claude Opus 4 and the efficient Claude Sonnet 4. These hybrid models introduce 'extended thinking' capabilities, allowing for deeper reasoning chains. Opus 4, in particular, is being hailed for its exceptional coding performance and ability to run autonomous AI agents for marathon sessions—reportedly up to seven hours straight—setting a new benchmark for complex, long-duration tasks. Early testers note its nuanced feedback on complex projects, with some users observing it providing insights that eluded both human reviewers and other leading models.
Claude Sonnet 4 is positioned as the new workhorse, offering near-instant responses and now serving as the base model for GitHub Copilot's coding agent, making it immediately accessible to millions of developers. The launch has sparked discussion about the future of software engineering, with some observers suggesting these models accelerate a shift from writing code to supervising AI-generated code. However, access is currently limited, with some API platforms like Windsurf noting they did not receive day-one access, highlighting ongoing capacity constraints. Pricing charts indicate Opus 4 remains a premium, expensive option for top-tier performance.
- Claude Opus 4 demonstrates superior coding and reasoning, capable of running autonomous AI agents for up to 7 hours.
- Claude Sonnet 4 is the new base model for GitHub Copilot's coding agent, offering fast, efficient performance.
- The launch highlights a strategic focus on AI agents and extended 'thinking' for complex, long-duration tasks.
Why It Matters
This release intensifies the AI coding assistant race, pushing capabilities toward autonomous agentic workflows that could reshape software development.