Anthropic Releases Claude Sonnet 4.5: The Best Coding Model Yet
Anthropic unveils Claude Sonnet 4.5, claiming the title of best coding model in the world with state-of-the-art performance on SWE-bench and significant improvements in reasoning, math, and agent capabilities.
Anthropic Releases Claude Sonnet 4.5: The Best Coding Model Yet
Anthropic has announced the release of Claude Sonnet 4.5, positioning it as the world's best coding model with groundbreaking improvements in reasoning, mathematics, and autonomous agent capabilities. The model launched on September 29, 2025, maintaining the same pricing as its predecessor while delivering substantial performance gains.
Breakthrough Performance
Claude Sonnet 4.5 has achieved state-of-the-art results on SWE-bench Verified, the industry standard for evaluating coding capabilities. The model also demonstrates exceptional computer use abilities, leading the OSWorld benchmark with an impressive 61.4% score – a significant jump from the previous 42.2%.
Key Improvements
The new model brings several major advancements:
- Extended Focus: Can maintain concentration on complex tasks for over 30 hours
- Enhanced Reasoning: Significant gains in logical reasoning and mathematical capabilities
- Superior Alignment: Described as the "most aligned frontier model" with reduced sycophancy and deception
- Improved Security: Better defense against prompt injection attacks
New Developer Tools
Alongside the model release, Anthropic introduced several powerful developer features:
Claude Code Enhancements
- Checkpoints: Save and restore conversation states for long-running projects
- Native VS Code Extension: Seamless integration with the popular code editor
API Improvements
- Context Editing: Modify conversation history without starting over
- Memory Tool: Persistent context across API sessions
Platform Expansions
- Code Execution in Claude Apps: Run and test code directly within the interface
- Claude for Chrome Extension: Browser-based assistance for web development
- Claude Agent SDK: Simplified framework for building autonomous agents
Pricing and Availability
Claude Sonnet 4.5 maintains the same pricing structure as Claude Sonnet 4 at $3 per million input tokens and $15 per million output tokens, making these substantial improvements accessible without additional cost.
Bonus Feature
Max subscribers can access a temporary "Imagine with Claude" research preview for five days following the release, offering early access to experimental image generation capabilities.
What This Means for Developers
Claude Sonnet 4.5 represents a significant leap forward in AI-assisted development. With its ability to handle complex, long-running tasks and its state-of-the-art coding performance, developers can tackle more ambitious projects with greater confidence. The enhanced agent capabilities make it particularly well-suited for building sophisticated automation tools and development workflows.
See It In Action
Want to compare how Sonnet 4.5 writes code as compared to Sonnet 4? Check out our code sample comparisons here.
Claude Sonnet 4.5 is now available through the Claude API and Claude Developer Platform, setting a new standard for AI-powered software development.