Anthropic has introduced Claude Sonnet 4.5, its most sophisticated AI model to date, which establishes a new standard in autonomous coding and application development. This model can operate continuously for 30 hours, generating fully functional applications with over 11,000 lines of code independently. This represents a significant improvement over its predecessor, Claude Opus 4.1, which was limited to about seven hours of autonomous operation. Anthropic reports that the new model successfully developed a chat platform similar to Slack and Microsoft Teams from scratch without interruption until completion. The company asserts that Claude Sonnet 4.5 is its most proficient model in utilizing computers, highlighting its commitment to creating advanced software agents.
This innovation builds on last year’s Computer Use feature, allowing AI to understand on-screen elements and engage with digital environments as a human would. In addition to its coding capabilities, Anthropic emphasizes the importance of alignment, noting that Sonnet 4.5 has been engineered to reduce problematic behaviors often found in advanced AI systems, such as sycophancy, deception, power-seeking, and promoting delusional thinking. Alongside Sonnet 4.5, Anthropic has enhanced its Claude Code and Claude API services, introducing a notable checkpoint feature that enables developers to revert projects to earlier code versions easily. For enterprise and advanced users, the Claude API now accommodates longer-running agents, which can tackle more intricate and multi-layered challenges.
Notably, Claude can now create slides, spreadsheets, and documents during conversations, broadening its application beyond mere coding to enhance workplace productivity. Anthropic has also launched a new creative tool, Imagine with Claude, which allows users to generate custom software in real-time from a single prompt, moving away from traditional programming frameworks. In a compelling demonstration, Claude was shown creating a complete desktop experience for William Shakespeare, including scripts, performance schedules, and related tools, showcasing the model’s ability to fuse creativity with practical application. For everyday users, Claude Sonnet 4.5 will be accessible for free, although daily usage may be limited by token restrictions. Premium subscribers will access the model at no extra cost, with subscription prices remaining unchanged.
Anthropic’s progress follows a recent OpenAI study indicating that Claude Opus 4.1 had already surpassed other AI agents, including GPT-5, approaching human-level performance in specific benchmarks. With the debut of Sonnet 4.5, expectations are even greater. For developers, businesses, and creatives, Claude Sonnet 4.5 may signify a pivotal moment in the role of AI in software development, productivity, and digital innovation.