Anthropic's Claude Sonnet 4.5 Shatters Focus Records: 30 Hours of Uninterrupted Multistep Task Completion

Anthropic Unveils Claude Sonnet 4.5: AI Model Maintains Focus for 30 Hours on Multistep Tasks On Monday, Anthropic released its latest language model, Claude Sonnet 4.5, which the company claims is its most capable model to date. According to Anthropic, Sonnet 4.5 has demonstrated a significant improvement in coding and computer use capabilities compared to previous models. The company's announcement also included the release of Claude Code 2.0, a command-line AI agent for developers, and the Claude Agent SDK, a tool that enables developers to build their own AI coding agents. Anthropic's CEO, Dario Amodei, stated in an interview that "the advancements we've made with Sonnet 4.5 are a major step forward in the development of more capable and reliable AI models." One of the key features of Sonnet 4.5 is its ability to maintain focus on complex tasks for extended periods. Anthropic reports that the model worked continuously on the same project for over 30 hours, a significant improvement compared to previous agentic models, which typically lose coherence over long periods due to errors and context windows filling up. In comparison, previous Claude 4.0 models were able to play Pokémon for over 24 hours or refactored code for seven hours. However, these tasks were relatively simple and did not require the same level of complexity as Sonnet 4.5's multistep tasks. To understand why Sonnet exists, it is essential to know how AI language models work. Traditionally, Anthropic has produced three differently sized AI models: small, standard, and large. These models are designed to handle various tasks, from simple text generation to complex coding and computer use capabilities. Anthropic's development of Sonnet 4.5 and its accompanying tools is part of a broader effort to create more capable and reliable AI models. The company's advancements in this area have significant implications for society, as they could lead to improved productivity, efficiency, and innovation in various industries. Experts in the field are cautiously optimistic about Anthropic's progress. Dr. Timnit Gebru, a leading researcher in AI ethics, noted that "while these developments are promising, it is essential to consider the potential risks and challenges associated with creating more capable AI models." As for next steps, Anthropic plans to continue refining its models and exploring their applications in various fields. The company's CEO emphasized the importance of collaboration between researchers, developers, and industry experts to ensure that AI models like Sonnet 4.5 are developed responsibly and safely. Anthropic's release of Claude Sonnet 4.5 marks a significant milestone in the development of more capable and reliable AI language models. As research continues to advance, it will be essential to consider the implications of these developments for society and ensure that they are used responsibly. *Reporting by Arstechnica.*