Anthropic has unveiled Opus 4.5, the final release in its 4.5 series of models, following the earlier launches of Sonnet 4.5 in September and Haiku 4.5 in October. The new version delivers top-tier performance across multiple benchmarks, including coding (SWE-Bench Verified and Terminal-bench), tool use (tau2-bench and MCP Atlas), and general problem solving (ARC-AGI 2 and GPQA Diamond). Notably, Opus 4.5 marks the first model to surpass 80% on SWE-Bench Verified, a widely respected benchmark for coding proficiency. The release highlights significant advancements in computer use and spreadsheet capabilities. To complement these upgrades, Anthropic is now making two new tools broadly available: Claude for Chrome and Claude for Excel. The Chrome extension will be accessible to all Max-tier users, while the Excel-focused version will be available to Max, Team, and Enterprise subscribers. Opus 4.5 also introduces enhanced memory management for long-context tasks, a critical improvement that required fundamental changes to how the model retains and retrieves information. Dianne Na Penn, Anthropic’s head of product management for research, emphasized that while longer context windows are helpful, the ability to remember the right details is equally vital. “Knowing the right details to remember is really important in complement to just having a longer context window,” she said. These improvements enable a new “endless chat” feature for paid Claude users. When the model reaches its context limit, it automatically compresses memory without interrupting the conversation, allowing seamless continuation without user alerts. Many of these enhancements are designed with agentic workflows in mind—particularly scenarios where Opus 4.5 acts as a lead agent coordinating multiple Haiku-powered sub-agents. Effective coordination demands strong working memory and the ability to navigate complex documents and codebases, including knowing when to backtrack or re-evaluate earlier decisions. Despite these advancements, Opus 4.5 enters a crowded market. It faces strong competition from recently launched frontier models, including OpenAI’s GPT-5.1, released on November 12, and Google’s Gemini 3, launched on November 18.