Anthropic introduces Claude Opus 4.5 with hybrid reasoning, benchmark gains, new Excel/Chrome tools

Anthropic has unveiled Claude Opus 4.5, positioning its flagship Claude model as a step up in coding, agentic workflows and everyday office tasks such as working with slides and spreadsheets.

The model has been released across Anthropic’s apps and API, with the company highlighting improvements in reasoning, computer use and long-horizon task execution.

Opus 4.5 is described by Anthropic as its most intelligent model to date, combining hybrid reasoning with a 200K-token context window.

Anthropic added the model will deliver state-of-the-art results on real‑world software engineering tests and stronger performance in multi‑step, tool‑using tasks typical of enterprise workflows.

What’s new in Opus 4.5

• Hybrid reasoning with an adjustable “effort” parameter for extended thinking when accuracy matters, balancing latency, cost and performance.
• 200K context window and improved memory handling for sustained, long‑context operations.
• "Endless chat" in the Claude apps, which allows paid users to continue conversations as the model compresses older context automatically.
• Broader product integrations: Claude for Chrome is available to all Max users; Claude for Excel is available to Max, Team and Enterprise users.
• Updates to the Claude Developer Platform and Claude Code to support longer‑running agents and more thorough planning and execution.

Benchmarks and cost

On Anthropic’s published evaluations, Opus 4.5 achieved 80.9% on SWE‑bench Verified and 66.3% on OSWorld for real‑world computer use, marking notable improvements over earlier Claude models.

Media overage also noted that Opus 4.5 is the first model to surpass 80% on SWE‑bench Verified.

Opus 4.5 is priced at $5 per million input tokens and $25 per million output tokens. It is available in Anthropic’s apps (Pro, Max, Team and Enterprise plans) and via the Claude API, and is accessible through Amazon Bedrock, Google Cloud’s Vertex AI and Microsoft Foundry.

For engineering teams, Anthropic has emphasised that Opus 4.5 can plan and deliver multi‑day software projects in hours, sustain autonomous coding sessions and operate more reliably as a lead agent that orchestrates sub‑agents.

For operations and finance users, Anthropic has pointed to stronger spreadsheet and document creation, alongside desktop and browser automation, to accelerate day‑to‑day workflows.