Anthropic’s new Opus 4.5 comes with Chrome and Excel integrations

Just a heads up, if you buy something through our links, we may get a small share of the sale. It’s one of the ways we keep the lights on here. Click here for more.

Anthropic just dropped a new flagship model and claimed it’s smarter, smoother, and somehow better at spreadsheets. 

Meet Opus 4.5, the final boss of Anthropic’s 4.5 lineup, following the earlier arrivals of Sonnet 4.5 in September and Haiku 4.5 in October. 

This is the last in the series, like the season finale, but with more benchmarks and fewer cliffhangers.

As expected, Opus 4.5 immediately started collecting trophies. 

It posted state-of-the-art scores across a laundry list of AI tests, including coding benchmarks like SWE-Bench and Terminal-Bench, tool-use evaluations like tau2-bench and MCP Atlas, and brain-melting problem-solving tests such as ARC-AGI 2 and GPQA Diamond. 

Its biggest flex? It’s the first model ever to score over 80% on SWE-Bench Verified, which is basically the AI equivalent of getting a gold star from every senior engineer on Earth.

But Anthropic isn’t just pitching Opus 4.5 as a smarter chatbot. They want it to be your new digital office assistant, one that never complains about meetings. 

Alongside the model, the company is expanding access to Claude for Chrome and Claude for Excel, two tools designed to show off Opus’ ability to handle browser tasks and spreadsheet chaos. 

The Chrome extension is heading to all Max users, while the Excel version will be available to Max, Team, and Enterprise customers, aka people who live inside rows and columns.

The real magic, though, is happening under the hood. Opus 4.5 features major memory upgrades for handling long conversations and massive documents. 

Instead of just stretching the context window, Anthropic focused on teaching the model what to remember. 

As head of product management, Dianne Na Penn put it, longer memory isn’t enough. The model needs to know which details actually matter.

These improvements also enabled a new “endless chat” feature for paid Claude users. 

When the model reaches its memory limit, it quietly compresses older parts of the conversation and keeps going, no awkward resets, no pop-up saying “sorry, we need to talk about your context window.”

All of this is especially important for Anthropic’s vision of agentic AI, where Opus 4.5 acts like a project manager, coordinating a team of smaller Haiku-powered sub-agents to tackle complex tasks. 

Think less “chatbot,” more “AI middle manager who never sleeps.”

Still, it’s entering a crowded battlefield, facing rivals like OpenAI’s GPT-5.1 and Google’s newly launched Gemini 3.

The AI arms race continues, now with better memory and fewer dropped conversations.