Anthropic releases Opus 4.5 with new Chrome and Excel integrations

On Monday, Anthropic introduced the latest iteration of its flagship AI model, Opus 4.5, marking the culmination of its 4.5 series. This release follows the earlier launches of Sonnet 4.5 and Haiku 4.5, both of which set high benchmarks in AI performance. Opus 4.5 showcases remarkable capabilities across various performance metrics, excelling in coding assessments like SWE-Bench and Terminal-bench, as well as tool utilization through benchmarks such as tau2-bench and MCP Atlas. A significant achievement for this new version is its ability to surpass 80 percent on the verified SWE-Bench, a prestigious coding evaluation. Additionally, Anthropic highlighted the model's enhanced functionality with computers and spreadsheets, unveiling parallel products to illustrate Opus 4.5’s proficiency in these environments. The Claude extensions for Chrome and Excel, previously in a pilot phase, are now being made widely accessible. The Chrome extension will roll out to all Max users, while the Excel integration will be available for Max, Team, and Enterprise users. Significant memory enhancements in Opus 4.5 allow for improved long-context operations, necessitating substantial adjustments in the model’s memory management. Dianne Na Penn, Anthropic’s head of product management for research, noted the importance of not just having longer context windows but also knowing which details to retain. This enables the new “endless chat” feature for paid Claude users, allowing uninterrupted conversations as the model compresses context memory without user notification. Many of these advancements are aimed at facilitating agentic applications, particularly where Opus serves as a primary agent directing a team of Haiku-powered sub-agents. Effective task management in these scenarios hinges on robust working memory, which the recent updates enhance significantly. Penn emphasized the critical role of memory in this context, stating that Claude must adeptly navigate codebases and extensive documents while being able to backtrack when necessary. As Opus 4.5 enters the competitive landscape, it will face challenges from other cutting-edge models, including OpenAI’s GPT 5.1 and Google’s Gemini 3, both of which have recently launched and are vying for dominance in the AI field.