SAN FRANCISCO—Anthropic, the Alphabet and Amazon.com-backed artificial intelligence startup, announced the launch of its newest model, Claude 4.5, on Monday. The release signals the company’s determined push into the lucrative enterprise AI sector, claiming the new version can handle longer, uninterrupted coding sessions and perform better on complex finance and scientific tasks.
Anthropic is in a fierce competition with rivals, notably OpenAI, to build the next generation of AI models that can reliably operate software and complete multi-step workflows. This capability is foundational for the development of AI agents—systems that can autonomously execute tasks on behalf of human users in a professional environment.
Chief Product Officer Mike Krieger highlighted the model’s advancements through internal testing. He stated that the new Sonnet 4.5 model successfully created a web application from scratch. Furthermore, one customer had the AI chatbot code autonomously for 30 hours, a significant leap from the seven-hour run Anthropic’s earlier flagship model, Claude Opus 4, achieved for a different client.
Krieger clarified that Anthropic is strategically targeting power users and business customers, prioritizing sustainable, reliable performance over fleeting consumer viral moments. Claude 4.5 shows marked improvement in several critical areas, including financial and scientific reasoning, and boasts a better ability to interact with computers. The company reported the model scored approximately 60% on a benchmark testing operating-system dexterity, a notable jump from the roughly 40% scored by prior iterations.
“It’s a lot more visceral when you just see the model using a computer the way a person does if you’re not a coder,” said Chief Science Officer Jared Kaplan, emphasizing the practical utility of the model’s enhanced dexterity.
In a separate but related development on Monday, Microsoft announced it would integrate new features into its Microsoft 365 Copilot powered by Anthropic models. These additions include “Agent Mode” in Excel and Word and an “Office Agent” within Copilot chat, with PowerPoint features soon to follow. This move by Microsoft, which was unveiled last week, aims to diversify the company’s AI ecosystem beyond its long-time partner, OpenAI.
Founded by former OpenAI executives, Anthropic has consistently positioned its Claude model for workplace use, implementing extensive guardrails designed to mitigate risky and unreliable outputs. The company has been aggressively marketing Claude’s advanced coding and data-analysis skills to highly regulated industries and enterprise teams that require models capable of working seamlessly across multiple software tools, underscoring its focus on sustained, reliable performance over long tasks rather than short, flashy demonstrations.
