Anthropic Unveils Opus 4.5 Featuring Chrome and Excel Tools
A new version of Opus has arrived, and it completes the 4.5 model series that rolled out over the past few months. The earlier releases included Sonnet 4.5 in September and Haiku 4.5 in October, and this latest addition continues that momentum. As expected, Opus 4.5 delivers state-of-the-art results across major benchmarks. Moreover, it excels in coding tests such as SWE-Bench and Terminal-bench, while also performing strongly on tool-use metrics like tau2-bench and MCP Atlas. It advances general reasoning too, scoring highly on ARC-AGI 2 and GPQA Diamond. Notably, the model is the first to surpass 80% on SWE-Bench verified.
The update also highlights better performance with computer-use tasks and spreadsheet workflows. Because of that focus, the release arrives alongside new tools designed to show how well the model handles real-world environments. With Opus 4.5, the Chrome extension becomes available to all Max users. Additionally, the spreadsheet-oriented version expands access to Max, Team, and Enterprise users. These additions broaden the model’s reach, and they also demonstrate how the system can support everyday productivity tasks.
Strengthening Long-Context Memory and User Experience
Opus 4.5 introduces refined long-context memory features, and these improvements required adjustments to how the model manages its internal processes. The upgrade enhances the system’s ability to retain important details rather than rely solely on extended context windows. This shift supports more accurate reasoning, smoother task continuity, and better handling of complex interactions.
Because of these changes, users now gain access to an “endless chat” feature. Instead of halting a conversation when the context window fills, the model compresses earlier messages. As a result, chats continue without interruption, and users avoid losing the flow of longer discussions. This improvement increases flexibility and creates a more natural interaction experience.
Preparing for Agentic Workflows and Growing Competition
Many enhancements support agentic workflows, especially scenarios where Opus directs smaller sub-agents powered by Haiku. Since these tasks rely on precise working-memory control, the new memory structure becomes especially valuable. The system can navigate code bases, examine large documents, and efficiently revisit earlier information when needed.
Even with these upgrades, Opus 4.5 enters a competitive landscape. Several advanced models launched this month, and both GPT 5.1 and Gemini 3 target similar use cases. However, Opus 4.5 arrives with strong performance indicators, smoother tools, and targeted improvements that position it firmly among today’s leading AI systems.
