Anthropic has released Opus 4.5, the newest version of its flagship model and the final addition to the company’s 4.5 lineup. The model achieves state-of-the-art performance across major benchmarks, including SWE-Bench, Terminal-bench, tau2-bench, MCP Atlas, ARC-AGI 2, and GPQA Diamond. It is also the first model to surpass 80 percent on SWE-Bench verified, a key coding benchmark.
The company highlighted improvements in computer-use, spreadsheet handling, and long-context operations. Alongside the model, Anthropic is expanding access to its Claude for Chrome and Claude for Excel tools, now available to Max users and to Team and Enterprise tiers for Excel.
Opus 4.5 incorporates substantial memory redesigns to improve long-context reliability and introduces an “endless chat” capability that compresses context seamlessly when limits are reached. These enhancements support agentic workflows in which Opus directs Haiku-based sub-agents across complex tasks such as software exploration and document analysis.
Anthropic positions Opus 4.5 against new frontier releases including OpenAI’s GPT 5.1 and Google’s Gemini 3, reflecting intensifying competition in advanced AI model development.
