AI BriefingClaude Sonnet 4.5
Key point
Anthropic has released Claude Sonnet 4.5, a model with greatly improved capabilities in coding, agent building, and computer use.
Details
Claude Sonnet 4.5 is a model that aims for world-class performance in coding, building complex agents, and computer use. It has also achieved significant advances in reasoning and math capabilities, and shows the highest level of alignment among the models Anthropic has released.
Major updates across the product ecosystem were also announced together:
- Claude Code: checkpoints feature that saves progress and allows instant rollback, a new terminal interface, and a native VS Code extension.
- Claude API: Added context editing and memory tool to support agents' long-running task execution.
- Claude Apps: Introduced code execution and file creation (spreadsheets, slides, documents) directly within the chat window.
- Claude Agent SDK: Provides tools that let developers build agents using Anthropic's infrastructure.
In terms of performance, Sonnet 4.5 recorded state-of-the-art results on SWE-bench Verified, and achieved 61.4% on the OSWorld benchmark, which measures real-world computer task capability, showing a dramatic improvement over the previous model (42.2%). Pricing remains the same as the existing Sonnet 4, at $3/$15 per million tokens.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.