This post is also available in:
עברית (Hebrew)
Anthropic has introduced Claude 4.5, the latest version of its large language model, designed to handle longer, more complex tasks with improved accuracy across domains like coding, finance, and scientific analysis. The release marks a continued move toward enterprise-focused AI systems capable of performing multi-step operations with minimal human oversight.
One of the key advancements in Claude 4.5 is its ability to sustain long coding sessions. In internal evaluations, the model was able to autonomously generate a functional web application and, in one client use case, continued coding uninterrupted for 30 hours—a significant jump from the 7-hour threshold achieved by the earlier Claude Opus 4.
Anthropic’s development of Claude 4.5 aligns with growing industry demand for AI agents that can reliably operate software, perform data analysis, and support decision-making across regulated sectors. The model scored around 60% on a benchmark assessing proficiency in operating-system-level tasks—an improvement over the 40% scored by previous iterations, According to Reuters. This suggests more advanced handling of complex workflows and better integration with existing digital tools.
Beyond core performance, Claude 4.5 also demonstrates stronger reasoning capabilities in finance and the sciences, positioning it as a useful assistant for industries with strict compliance and analytical demands.
In a related move, Microsoft announced new features in its 365 Copilot suite powered by Anthropic models. These include additions like “Agent Mode” in Excel and Word, and an “Office Agent” in Copilot chat, with support for PowerPoint expected to follow. The integration signals Microsoft’s intent to diversify its AI backend beyond its reliance on OpenAI.
Claude’s design emphasizes consistency and safety, aiming to reduce problematic outputs while maintaining performance over extended tasks. This approach is particularly relevant for enterprise users looking to deploy AI across business operations that require reliability.
With Claude 4.5, Anthropic continues to develop models geared toward real-world applications—especially those requiring sustained attention, data handling, and software interaction across organizational environments.
