Claude Opus 4.5 Launches With Wider Chrome and Excel Access

Anthropic has released Claude Opus 4.5, the latest version of its flagship AI model. The launch completes the company’s 4.5 lineup and brings major upgrades across performance, memory, and real-world tool use. The release also comes with wider access to Anthropic’s Chrome and Excel integrations, which were previously available to selected users only.

Performance jumps across key benchmarks

The new version of Opus sets fresh records on several respected benchmarks. It crosses 80% on SWE-Bench verified, which represents a significant step forward for coding accuracy. It also posts strong results on tool-use tests like tau2-bench and MCP Atlas, and shows improved performance in ARC-AGI 2 and GPQA Diamond general reasoning benchmarks.

Anthropic highlights how Opus 4.5 handles complex coding tasks, multi-step problem solving, and agent-driven operations more reliably. The company says the model is designed to perform as a lead agent in workflows supported by sub-agents powered by Haiku.

Chrome and Excel integrations get wider availability

Along with Opus 4.5, Anthropic is expanding access to two important integrations. The Claude for Chrome extension is now available for all Max users. Meanwhile, Claude for Excel rolls out to Max, Team, and Enterprise users.

These tools allow users to work directly inside browsers and spreadsheets without switching apps. Anthropic positions this as a move to make AI-assisted software use more natural and less disruptive.

New memory system enables ‘endless chat’

Opus 4.5 also comes with improvements in long-context memory. Anthropic says the upgrades required fundamental changes to how the model stores and retrieves information. The result is a more stable experience when handling long documents, large codebases, and multi-step instructions.

One of the most requested features — “endless chat” — is finally available for paid Claude users. When conversations reach the context limit, the model compresses older memory without interrupting the session.

Facing tough competition in the frontier model race

The launch of Opus 4.5 arrives during an intense wave of AI releases. The new model joins the market alongside OpenAI’s GPT 5.1 and Google’s Gemini 3, both rolled out earlier this month.

However, Anthropic is betting on performance consistency, tool-use reliability, and stronger memory to help Opus 4.5 stand out in real-world usage.