Anthropic has officially launched its most advanced AI model yet, Claude Sonnet 4.5, and it is already creating headlines in the tech world. The company revealed that the model can autonomously code for 30 continuous hours, completing full applications from scratch. In its most impressive demonstration, Claude Sonnet 4.5 built a chat app similar to Slack and Microsoft Teams with more than 11,000 lines of code, stopping only after the task was fully accomplished. This marks a giant leap forward compared to its predecessor Opus 4.1, which could only sustain coding for up to seven hours.

Claude Sonnet 4.5 extends Anthropic’s “Computer Use” feature, enabling the AI to interpret screens, navigate tasks, and function more like a human developer. The company emphasizes that this model is its most aligned version yet, having significantly reduced problematic tendencies such as sycophancy, deception, and power-seeking behaviors. It also integrates updates to Claude Code, including a new checkpoint feature that allows developers to roll back to previous versions easily.

Alongside Sonnet 4.5, Anthropic has updated the Claude API to allow agents to run for longer periods and tackle more complex challenges, including creating files such as slides, spreadsheets, and documents directly during chats. A standout addition is “Imagine with Claude,” which lets users generate software instantly based on a single prompt. In a live demo, Claude even created a Shakespeare-themed desktop environment complete with play scripts and schedules. Free users will gain limited access to Claude Sonnet 4.5, while premium users will enjoy unrestricted performance without any price hike.

Industry experts believe that Claude Sonnet 4.5 could outpace earlier benchmarks set by Claude Opus 4.1, which had already rivaled GPT-5 in certain tasks. With its enhanced capabilities, Claude Sonnet 4.5 positions itself as one of the most powerful AI coding and creative tools available in 2025, blending efficiency, autonomy, and human-like adaptability.