Anthropic Unveils Claude Sonnet 4.5 with Enhanced Programming

Today, Anthropic has introduced a new model — Claude Sonnet 4.5, which, according to the developers, is one of the best for programming in the world, surpassing GPT-5 and Gemini 2.5 Pro. It is one of the most powerful models for creating complex agents and working with computers. Anthropic also notes significant improvements in reasoning and mathematical abilities compared to Opus 4.1.

Along with this, the company has added a new terminal and checkpoints to Claude Code. Checkpoints allow users to save progress and return to previous states when necessary. Now, Claude applications support code execution and file creation for spreadsheets, presentations, and documents directly during conversations.

Users can try Claude for Chrome, as Anthropic is allowing some users to test this feature in beta mode. An SDK is also available for creating agents, opening up additional opportunities for developers.

According to the company, Claude Sonnet 4.5 has enhanced capabilities and a wide range of safety training. This helps reduce sycophancy, deception, power-seeking, and tendencies to encourage illusory thinking. The new model is available everywhere starting today, and Anthropic recommends users upgrade to the new version.