ANTHROPIC
Opus 4.5 just showed everyone how memory should work
Anthropic has launched Opus 4.5, the final model in its 4.5 lineup after Sonnet and Haiku.
It hits top scores across coding, tool use, and reasoning tests, and it’s the first model to pass 80% on SWE-Bench Verified, a major coding benchmark.
The Chrome extension will be available to Max users, while the Excel version will reach Max, Team, and Enterprise plans.
A big part of this update is better memory.
Opus 4.5 can handle longer documents and multi-step tasks more smoothly, and it introduces an “endless chat” option for paid users.
Three things to know:
Opus 4.5 focuses on stronger long-context memory, not just bigger context windows.
Claude for Chrome and Excel are rolling out widely.
The model is built for agentic, multi-step workflows.
Built for big-brain tasks
Instead of stopping when it reaches the context limit, it quietly compresses older details and keeps going.
These upgrades are designed for agent-style work, where Opus can guide smaller Haiku-based agents through tasks like reading codebases or checking long documents.
Opus 4.5 now enters a crowded field, launching alongside OpenAI’s GPT-5.1 and Google’s Gemini 3.
We need an AI cage match to determine the winner. - MV