Claude Sonnet 4.5: Anthropic's Latest AI Model Release

Anthropic has officially released Claude Sonnet 4.5, a significant upgrade to their AI model, boasting state-of-the-art performance in coding and complex agentic applications. This new version introduces a range of enhancements designed to make it a more powerful and versatile tool for developers and users alike.

Enhanced Coding and Agent Capabilities

Sonnet 4.5 demonstrates substantial advancements in its ability to handle coding tasks, from generating and analyzing code to identifying optimal improvements and refactoring. It particularly excels at autonomous long-horizon coding tasks, where it can plan and execute complex software projects over extended periods. The model's improved tool handling, memory management, and context processing contribute to its enhanced agent capabilities.

Performance Benchmarks

Claude Sonnet 4.5 has shown significant performance gains on industry-standard benchmarks:

- OSWorld: The model achieves a score of 61.4% on this test for real-world computer tasks, demonstrating its ability to navigate and interact with operating systems and applications.

- SWE-bench Verified: With a score of 77.2%, Sonnet 4.5 excels on this benchmark for software engineering tasks, showing substantial improvement in real-world coding scenarios.

These results position Claude Sonnet 4.5 as a leading model for practical coding and agentic applications.

Improved Safety Profile

Anthropic has focused significantly on improving the model's safety characteristics. Claude Sonnet 4.5 shows reduced rates of concerning behaviors including sycophancy and deception. The company has also made considerable progress in defending against prompt injection attacks, a critical security concern for agentic capabilities. Additionally, improvements have been made to prevent the model from encouraging delusional thinking or reinforcing unhealthy patterns.

Pricing and Availability

Sonnet 4.5 is a drop-in replacement for its predecessor, offering improved performance at the same price point of $3 per million input tokens and $15 per million output tokens. The model is available through:

- Anthropic's web, mobile, and desktop chat interfaces

- The Claude API and developer platform

- Claude Code, a command line tool for agentic coding

- Amazon Bedrock, a fully managed service offering access to foundation models

With its combination of enhanced capabilities, strong benchmark performance, and improved safety features, Claude Sonnet 4.5 represents a significant step forward in AI model development for practical applications.

Written by Claude