Anthropic unveils Claude Sonnet 4.5, claiming premier AI for agentic coding

In the escalating generative AI arms race, Anthropic has launched Claude Sonnet 4.5 that sets a new benchmark for productivity and reliability in software development.

The model was released on Monday, September 29, within two months after the launch of Claude Opus 4.1, reflecting an increasing acceleration of model updates and directly challenging OpenAI's new update GPT-5.

The company states that Sonnet 4.5 is not just an incremental update to the previous model but offers a foundational leap in creating “production-ready” applications.

According to Anthropic's official website, “Claude Sonnet 4.5 is the best coding model in the world. It's the strongest model for building complex agents.”

Claude Sonnet 4.5 is also SWE-Bench verified, analysing real-world software engineering abilities by resolving actual GitHub issues.

Technical prowess and autonomous operation

The internal testing of the model was conducted by Anthropic researcher David Hershey. He reported that Claude Sonnet 4.5 operated autonomously for over 30 hours on a single complex task, showcasing remarkable endurance.

Further technical enhancements include:

• On the OSWorld benchmark, which tests AI models on real-world computer tasks, Sonnet 4.5 achieved a leading score of 61.4%, a substantial gain from Sonnet 4's 42.2% just four months prior.
• Anthropic emphasizes that this is its "most aligned frontier model" to date, with reduced instances of flattery, false claims, and improved resistance to prompt injection attacks.

Ecosystem expansion with the Claude agent SDK

Alongside the model, Anthropic is democratizing the infrastructure behind its own products by releasing the Claude Agent SDK. This toolkit provides developers with the same building blocks used to create "Claude Code," allowing for the construction of sophisticated, custom AI agents. This move is strategic, empowering a developer ecosystem to build upon Anthropic's technology and cementing its platform's utility.

The product suite also receives major upgrades, including:

• Checkpoints in Claude Code: A highly-requested feature that saves progress and allows for instant rollbacks.
• Enhanced API Features: New context editing and memory tools enable agents to handle greater complexity over longer runtimes.
• Native Integrations: A new native VS Code extension and the release of the Claude for Chrome extension for Max subscribers, which showcases the model's ability to navigate browsers and manipulate applications like spreadsheets in real-time.

The release of Claude Sonnet 4.5 intensifies the pressure on competitors and marks a pivotal step towards AI as a collaborative engineer.