Claude Sonnet 4.5 – Anthropic’s latest AI programming model

What is Claude Sonnet 4.5?

Claude Sonnet 4.5 is Anthropic’s latest and most powerful programming model. It demonstrates outstanding performance in programming, computer operations, reasoning, and mathematics. It ranked at the top of the SWE-bench Verified benchmark and can stay focused on tasks for over 30 hours. The model features advanced agentic capabilities, enabling complex task automation via APIs and tools. Claude Sonnet 4.5 also introduces significant improvements in alignment and safety, reducing harmful behaviors and enhancing defense mechanisms. It supports a wide range of applications including code generation, knowledge Q&A, and content creation. Currently, Claude Sonnet 4.5 is fully available on the claude.ai website, app, and API (with the same pricing as its predecessor, Claude Sonnet 4).

Claude Sonnet 4.5 also introduces the “Imagine with Claude” temporary research preview, which allows users to interact with Claude using natural language instructions to generate and adjust software code and functionalities in real time. All output is instantly created without requiring predefined code. This feature is currently exclusive to Max subscribers, designed to accelerate software development, prototyping, and creative exploration with a highly efficient and flexible interactive experience.

Main Features of Claude Sonnet 4.5

• Long-running agents: Provides excellent instruction following, tool use, error correction, and advanced reasoning, suitable for customer-facing agents and complex AI workflows.

• Code generation: Supports the full software development lifecycle, from planning to bug fixing, maintenance, and large-scale refactoring, with up to 64K output tokens—ideal for complex code generation and planning.

• Browser and computer operations: Excels at handling computer-based tasks, reliably performing browser-driven workflows such as competitive analysis, procurement processes, and customer onboarding, with improved accuracy over previous versions.

• Cybersecurity: Teams using Sonnet 4.5 and Claude Code can deploy agents to automatically patch vulnerabilities before exploitation, shifting from passive detection to proactive defense.

• Financial analysis: Handles everything from basic financial reporting to advanced predictive analytics, such as continuously monitoring global regulatory changes and proactively adapting compliance systems.

• Business productivity: Strong at creating and editing presentations, documents, and spreadsheets.

• Research: Can query both external and internal data sources, synthesizing comprehensive insights across complex information environments.

• Content generation and analysis: Produces high-quality writing with nuanced understanding of tone and context, generating engaging content and performing in-depth analysis.

Core Upgrades in Claude Sonnet 4.5

• Claude Code: Introduces checkpointing, allowing users to save progress and roll back instantly, along with an updated terminal interface and a native VS Code extension to improve developer experience.

• Claude API: Adds context editing and memory tools, enabling agents to run longer and handle more complex tasks, expanding functionality and use cases.

• Claude App: Integrates code execution and file creation (spreadsheets, slides, and documents) directly into conversations. Max users who joined last month’s waitlist now also have access to a Chrome extension for greater convenience.

• Claude Agent SDK: Provides developers with the same infrastructure and tools used to build Claude Code, empowering them to create their own cutting-edge products.

Performance of Claude Sonnet 4.5

• SWE-bench Verified: Achieved industry-leading results. In practice, the model can stay engaged with complex, multi-step tasks for over 30 hours.

• OSWorld Benchmark: Reached 61.4% accuracy, a significant improvement over prior models.

• Chrome browser extension: Works directly inside the browser via Claude for Chrome, enabling tasks such as browsing websites, filling spreadsheets, and completing workflows.

• Reasoning and mathematics: Shows improved abilities compared to previous models, excelling in general reasoning and advanced math benchmarks.

• Domain-specific expertise: Experts in finance, law, medicine, and STEM report significant improvements in specialized knowledge and reasoning compared with earlier models, including Opus 4.1.

Application Scenarios of Claude Sonnet 4.5

• Long-running agents: Ideal for tasks requiring extended runtime, such as workflow automation, continuous monitoring, and multi-step processes.

• Code generation: Supports the entire software development lifecycle, from planning to implementation, maintenance, and refactoring.

• Browser and computer operations: Automates web-based tasks, including data extraction, process management, and online workflows.

• Cybersecurity: Deploys agents to proactively detect and fix vulnerabilities, shifting from reactive defense to proactive security.

• Financial analysis: Performs tasks from basic financial evaluations to advanced compliance monitoring and predictive analytics.

Who Should Use Claude Sonnet 4.5

• Software developers: Take advantage of powerful code generation and optimization to accelerate complex programming tasks.

• Cybersecurity teams: Automatically detect and patch vulnerabilities, moving from passive defense to proactive protection.

• Financial analysts: Conduct both basic and advanced financial analysis, including risk assessment and compliance monitoring, supporting intelligent risk management.

• Enterprise professionals: Efficiently create and edit office documents such as slides, reports, and spreadsheets, improving daily productivity.

• Researchers: Integrate external and internal data sources to generate comprehensive insights across complex domains for academic and business research.