My Honest Impression of Claude Sonnet 4.5
I tried Claude Sonnet 4.5, which was released on September 29, 2025.
In conclusion, the experience was excellent.
While the chat functionality has improved, the coding capabilities have truly reached an outstanding level.
The Cutting Edge of Rapidly Evolving AI Models
AI technology never stops evolving.
Claude Sonnet 4.5 represents the latest achievement, delivering significant performance improvements over previous models.
I’m currently using Claude Sonnet 4.5 as my main model.
Official Recognition as the World’s Best Coding Model
Anthropic officially announced it as “the best coding model in the world.”
This isn’t just marketing hype.
It achieved a 77.2% score on SWE-bench Verified, objectively proving its real-world software development capabilities.
With advanced parallel test-time compute, it can reach up to 82.0%.
Remarkable Focus: Over 30 Hours of Continuous Work
The most impressive feature is its endurance.
Claude Sonnet 4.5 can maintain focus on complex, multi-step tasks for over 30 hours.
It offers the reliability needed for long-duration development work such as large-scale refactoring and feature additions.
Key Features of Claude Sonnet 4.5
Claude Sonnet 4.5 includes the following officially announced features:
Significant Coding Performance Improvements
It achieved a 77.2% score on SWE-bench Verified.
This is the average of 10 trials, reaching 82.0% with advanced parallel test-time compute.
Code generation accuracy has improved, and refactoring judgments have become more appropriate.
Security measures and bug detection capabilities have also been enhanced.
Revolutionary Agent Capabilities
Extended autonomous operation is now possible.
Tool handling, memory management, and context processing have been significantly improved.
On the OSWorld benchmark, it scored 61.4%, a dramatic improvement from Claude Sonnet 4’s 42.2% just four months ago.
New API Features
The context editing feature automatically clears information from old tool calls.
The memory tool enables storing and referencing information outside the context window.
These features allow for more efficient processing of longer and more complex tasks.
Maintained Cost-Performance
Pricing remains the same as Claude Sonnet 4.
Available at $3 input, $15 output per million tokens.
Despite significant performance improvements, the unchanged pricing is a major advantage.
Enhanced Performance in Specialized Domains
Performance has dramatically improved in finance, law, medicine, and STEM fields.
Expert evaluations show superior domain-specific knowledge and reasoning capabilities compared to Claude Opus 4.1.
Extended Thinking mode is supported, enabling more complex reasoning.
Practical Use Cases
Claude Sonnet 4.5 excels in various scenarios.
In development work, it handles everything from code generation to refactoring and bug fixes.
Early access users have reported adoption in development tools like Cursor and GitHub Copilot, contributing to solving complex problems.
In cybersecurity, early adopters reported reducing average vulnerability intake time by 44% and improving accuracy by 25%.
In financial analysis, it provides investment-grade insights for risk analysis, structured products, and portfolio screening.
In legal work, it delivers state-of-the-art performance on the most complex tasks, such as analyzing litigation records and drafting judicial opinions.
The Most Aligned Model
Claude Sonnet 4.5 is Anthropic’s most aligned frontier model to date.
Concerning behaviors such as sycophancy, deception, power-seeking, and encouraging delusional thinking have been significantly reduced.
Defense against prompt injection attacks has also been substantially improved.
Summary: Why I Adopted It as My Main Model
For these reasons, I’m using Claude Sonnet 4.5 as my main model.
It truly demonstrates its value especially in coding work.
The world-class performance proven by SWE-bench Verified is a tremendous help in actual development environments.
The ability to work continuously for over 30 hours provides confidence even for large-scale projects.
AI technology continues to accelerate its evolution.
Claude Sonnet 4.5 is a revolutionary model at the forefront of this progress.
I highly recommend developers experience this remarkable coding capability for themselves.