anthropic releases opus 4.5 with new chrome and excel integrations

Anthropic Unveils Opus 4.5: A New Frontier in AI Capabilities

Anthropic officially introduced Opus 4.5 on Monday, representing the newest iteration of its leading model. This release completes Anthropic’s 4.5 series, following the prior launches of Sonnet 4.5 in September and Haiku 4.5 in October.

Exceptional Performance Across Key Benchmarks

The updated Opus demonstrates cutting-edge performance across a diverse array of benchmarks. These include assessments for coding proficiency (SWE-Bench and Terminal-bench), tool utilization (tau2-bench and MCP Atlas), and comprehensive problem-solving skills (ARC-AGI 2, GPQA Diamond).

Significantly, Opus 4.5 is the inaugural model to surpass an 80% score on the highly regarded SWE-Bench verified coding benchmark.

Enhanced Computer Use and Spreadsheet Integration

Anthropic highlighted Opus’ enhanced capabilities in computer operation and spreadsheet management. To demonstrate these strengths, several complementary products are being launched alongside the model.

Alongside Opus 4.5, Anthropic is expanding the availability of Claude for Chrome and Claude for Excel, which were previously available in a pilot program. The Chrome extension will be accessible to all Max subscribers. The Excel integration will be offered to Max, Team, and Enterprise users.

Improvements to Long-Context Memory

Opus 4.5 incorporates notable advancements in memory management for extended-context operations. These improvements necessitated substantial modifications to the model’s memory handling processes.

“We’ve implemented improvements to general long context quality during training with Opus 4.5, but simply increasing context windows isn’t enough,” explained Dianne Na Penn, Anthropic’s head of product management for research, in an interview with TechCrunch. “The ability to retain the most relevant information is crucial, in addition to having a larger context window.”

Introducing “Endless Chat” for Seamless Conversations

These memory enhancements have also enabled the implementation of a long-awaited “endless chat” feature for paying Claude users. This feature allows conversations to continue uninterrupted when the model reaches its context window limit.

The model will now compress its context memory automatically, without notifying the user.

Optimized for Agentic Workflows

Many of these upgrades are geared towards agentic applications, particularly scenarios where Opus functions as a primary agent directing a team of Haiku-powered sub-agents. Effective management of these tasks requires robust working memory, which is where the described memory improvements prove invaluable.

“Fundamental aspects like memory are particularly important in these scenarios,” Penn stated. “Claude needs to be able to navigate codebases and extensive documents, and also know when to revisit and verify information.”

Competition in the Frontier Model Landscape

Opus 4.5 will encounter strong competition from other recently released advanced models. Notable contenders include OpenAI’s GPT 5.1 (released November 12) and Google’s Gemini 3 (released November 18).

• Opus 4.5: Anthropic’s latest flagship model.
• Sonnet 4.5 & Haiku 4.5: Previous releases in the 4.5 series.
• SWE-Bench: A respected coding benchmark.