Anthropic on Monday unveiled Opus 4.5, the final and most powerful model in its 4.5 lineup, following earlier releases of Sonnet and Haiku.
The company says Opus 4.5 delivers state-of-the-art results across coding, tool use, and general reasoning tests, becoming the first model to break 80% on the respected SWE-Bench verified coding benchmark.
To highlight its improved computer-use and spreadsheet skills, Anthropic is also expanding access to Claude for Chrome and Claude for Excel, which will now roll out more broadly to paid users.
Opus 4.5 introduces major upgrades to long-context memory, enabling an “endless chat” feature that quietly compresses past conversation so users no longer hit context limits.
These memory gains are designed to support increasingly agentic workflows, where Opus leads teams of smaller Haiku-based agents through complex tasks like navigating large codebases.
The model enters a tight race against other top-tier systems, including OpenAI’s newly released GPT 5.1 and Google’s Gemini 3.
