Anthropic rolls out Claude Opus 4.5 and new Claude Code desktop app

Anthropic has launched Claude Opus 4.5 with improved coding, reasoning and long-form task performance, alongside a new Claude Code desktop app for PC and Mac

Anthropic has introduced Claude Opus 4.5, a new flagship AI model designed to improve software development, reasoning and complex multi-step work. The model is now available via the Claude apps, API and select cloud platforms. It will compete with other models like OpenAI’s GPT 5.1 and Google’s Gemini 3. Additionally, Anthropic has released the Claude Code desktop app for Windows, macOS and Windows (Arm 64).

Claude Opus 4.5: What’s new

According to Anthropic, Opus 4.5 delivers its biggest gains in software engineering tasks, including debugging and working with multi-system codebases, when compared with older models. Internal testing cited by the company suggests that the model can handle more ambiguity and technical trade-offs and can complete tasks that earlier models struggled with. Beyond programming, the model is also designed to perform better in document-heavy scenarios such as analysing slides, spreadsheets and long research tasks.

 

Anthropic says Opus 4.5 improves on several capability areas, including vision, mathematics and multi-step reasoning. The company highlighted examples where the model found workable solutions in test environments that exceeded the expectations of standard benchmarks, such as resolving an airline itinerary issue by identifying an alternative procedural path.

The company also claims stronger protection against prompt-injection attacks, a method used to trick AI systems into ignoring instructions. Opus 4.5 is described as its most robustly aligned model to date, with upgrades intended to better detect and avoid harmful or deceptive inputs.

Anthropic product updates

The Opus 4.5 AI model launch is accompanied by updates across Anthropic’s product ecosystem. According to the company, Claude Code gains more structured planning tools and is now available in the desktop app, allowing multiple coding or research sessions to run in parallel.

 

For general users, the Claude apps now handle longer conversations by automatically summarising earlier context. Claude for Chrome is expanding to more subscribers, and the company has widened beta access to Claude for Excel for Max, Team and Enterprise users. Anthropic has also adjusted usage limits so that users with access to Opus 4.5 can run it at levels similar to the previous Sonnet tier.