Anthropic Unveils Claude Sonnet 4.5, Elevating AI Coding and Agent Creation

September 29, 2025

Details

  • Anthropic introduced Claude Sonnet 4.5 on September 29, 2025, touting it as their top-performing coding model, with state-of-the-art results on the SWE-bench Verified benchmark and the ability to sustain focus on complex projects for more than 30 hours.
  • The update brings significant enhancements to Claude Code, now supporting progress-saving checkpoints, a built-in VS Code extension, a revamped terminal interface, and new context editing tools for extended agent-driven tasks.
  • Claude Sonnet 4.5 scored 61.4% on the OSWorld computer-use benchmark, a substantial leap from its predecessor's 42.2%, highlighting notable advancements in performing real-world coding and computational tasks.
  • The launch also debuts the Claude Agent SDK, offering developers access to the underlying infrastructure that powers Claude Code, enabling custom AI agent development for a range of applications beyond coding.
  • Pricing remains consistent with the previous version at $3 and $15 per million tokens, as Anthropic frames this as their most reliably aligned model yet, reducing risks like sycophancy and deceptive behavior.

Impact

The release positions Anthropic as a formidable rival to OpenAI's Codex and GitHub Copilot, underscoring rapid innovation in AI-driven software development tools. With its improved agent capabilities and developer tools, Claude Sonnet 4.5 may accelerate adoption of autonomous AI in businesses. Anthropic's focus on alignment and real-world utility sets a new bar for trust and performance in enterprise AI applications.

Read more