Anthropic has launched the flagship model, Opus 4.5, which delivers a cutting-edge performance across major benchmarks, including coding (SWE-Bench, Terminal-bench), tool use (tau2-bench, MCP Atlas) and general reasoning (ARC-AGI 2, GPQA Diamond).
The recently introduced model has scored over 80% on SWE-Bench verified, marking a significant milestone in coding capability.
The company emphasised Opus 4.5’s enhancements in computer use and spreadsheet-based tasks.
Apart from this model, Anthropic is expanding access to Claude for Chrome and Claude for Excel. The Chrome Extension will be accessible to all Max users, while the Excel tool will be available to Max, Team and Enterprise customers.
The update is primarily focused on long-context memory improvement, needing large architectural changes.
These upgrades power a new “endless chat” experience for paid Claude users. When the model reaches its context limit, it will compress past conversation history automatically.
The improved memory system supports agentic workflows, where Opus leads several tasks implemented by Haiku-based sub-agents. Users can navigate large codebases or documents and manage working memory efficiently.
Opus 4.5 launch comes shortly after the release of OpenAI’s GPT 5.1 and Google’s Gemini 3.
