Anthropic has announced the release of its most advanced AI model yet, Claude Sonnet 4.5, setting a new benchmark in autonomous coding and app development. The company revealed that the model is capable of working continuously for 30 hours, creating fully functional applications with more than 11,000 lines of code written independently.

This marks a significant leap from its predecessor, Claude Opus 4.1, which could only sustain autonomous work for around seven hours. According to Anthropic, the new model successfully built a chat platform resembling Slack and Microsoft Teams—from scratch—without halting until the project was complete.

Anthropic claims that Claude Sonnet 4.5 is its “best model at using computers,” emphasizing its focus on building sophisticated software agents. The innovation builds upon the company’s Computer Use feature introduced last year, which enables AI to interpret on-screen elements and interact with digital environments much like a human operator.

Beyond raw coding power, Anthropic highlights another achievement: alignment. The company insists Sonnet 4.5 has been designed to minimize concerning behaviors often associated with advanced AI systems, such as “sycophancy, deception, power-seeking, and the tendency to encourage delusional thinking.”

Updates to Claude Code and Claude API

Alongside Sonnet 4.5, Anthropic has rolled out improvements to its Claude Code and Claude API services. One standout addition is the checkpoint feature, allowing developers to roll back projects to earlier code versions with ease.

For enterprise and advanced users, the Claude API now supports longer-running agents, enabling them to solve more complex and multi-layered problems. Impressively, Claude can now generate slides, spreadsheets, and documents directly during conversations, expanding its utility beyond code to workplace productivity.

Imagine with Claude: Creativity Reimagined

Anthropic has also unveiled a new creative tool called Imagine with Claude. Unlike traditional programming frameworks with preset commands, this feature empowers users to generate bespoke software on the fly from a single prompt.

In a striking demonstration, Anthropic showed Claude designing an entire desktop experience for William Shakespeare, complete with scripts, performance schedules, and related utilities—illustrating how the model can blend imagination with functionality.

Access and Pricing

The good news for everyday users is that Claude Sonnet 4.5 will be available for free, though daily usage will likely be capped by token limits. Premium subscribers will gain access to the model at no additional cost, as Anthropic has confirmed that subscription prices remain unchanged.

Performance Beyond Expectations

Anthropic’s advancements come shortly after a study by OpenAI revealed that Claude Opus 4.1 had already outperformed other AI agents, including GPT-5, and had come close to human-level performance in certain benchmarks. With the release of Sonnet 4.5, expectations are even higher.

For developers, businesses, and creatives alike, Claude Sonnet 4.5 could represent a major turning point in how AI contributes to software development, productivity, and digital innovation.