Anthropic releases Opus 4.5 with new Chrome and Excel integrations (techcrunch.com)

Anthropic announced Opus 4.5, the final model in its 4.5 series (following Sonnet and Haiku 4.5), touting state-of-the-art results across coding, tool use, and general reasoning benchmarks — including SWE‑Bench, Terminal‑bench, tau2‑bench, MCP Atlas, ARC‑AGI 2 and GPQA Diamond. Notably, Opus 4.5 is the first model to score over 80% on SWE‑Bench verified, signaling a material boost in coding accuracy. Anthropic is also broadening real‑world integrations: Claude for Chrome will roll out to all Max users, and Claude for Excel will become available to Max, Team and Enterprise customers. Technically, Opus 4.5 introduces significant memory‑management changes to improve long‑context quality, enabling features like an “endless chat” for paid Claude users that compresses context when windows fill rather than truncating conversations. Those memory advances are aimed at agentic workflows — e.g., Opus acting as a lead agent coordinating Haiku‑powered subagents and navigating large codebases or documents while deciding what to remember and when to backtrack. The release tightens competition at the frontier with OpenAI’s GPT‑5.1 and Google’s Gemini 3, and underscores a broader shift toward models optimized for tool use, sustained multi‑step work, and integrated desktop/browser assistants.