Anthropic has officially launched Claude Sonnet 4.5, a new frontier model that the company asserts delivers state-of-the-art performance in coding benchmarks. This latest iteration is designed to move beyond mere prototypes, offering the capability to construct “production-ready” applications, signaling a significant advancement in reliability compared to its predecessors. Claude Sonnet 4.5 is immediately accessible to users through the Claude API and within the Claude chatbot platform.

For developers, the pricing structure for Claude Sonnet 4.5 remains consistent with that of Claude Sonnet 4, set at $3 per million input tokens—an amount equivalent to approximately 750,000 words, surpassing the entire text of the Lord of The Rings series—and $15 per million output tokens. Over the past year, Anthropic’s AI models have gained considerable traction among developers and enterprises, largely due to their robust performance in diverse software engineering tasks. Notably, major tech entities such as Apple and Meta reportedly utilize Claude AI models internally, while Anthropic has established a substantial business by providing API access to various AI coding applications, including Cursor, Windsurf, and Replit.

The competitive landscape in the AI sector has seen recent shifts, with OpenAI’s GPT-5 emerging as a strong contender, challenging Anthropic’s established dominance and reportedly outperforming Claude models on several coding benchmarks. Despite this, Anthropic maintains that Claude Sonnet 4.5 achieves industry-leading performance on various coding benchmarks, including SWE-Bench Verified. However, Anthropic AI researcher David Hershey highlighted to TechCrunch that the full scope of Claude Sonnet 4.5’s performance extends beyond what benchmarks alone can capture.

Hershey shared insights from early trials with enterprise customers, where he observed Claude Sonnet 4.5 coding autonomously for extended periods, up to 30 hours. During these trials, the AI model demonstrated remarkable capabilities, not only building applications but also autonomously setting up database services, acquiring domain names, and conducting SOC 2 audits to ensure product security. Industry leaders have echoed these sentiments; Cursor CEO Michael Truell affirmed that Claude Sonnet 4.5 represents state-of-the-art coding performance, particularly in managing longer-horizon tasks, while Windsurf CEO Jeff Wang praised it as a “new generation of coding models.”

Beyond its technical prowess in coding, Anthropic also claims that Claude Sonnet 4.5 is its most aligned frontier AI model to date. The company reports reduced instances of sycophancy and deception compared to earlier models, alongside enhanced resilience against prompt injection attacks, underscoring a commitment to ethical AI development and security. Concurrently with the launch of Claude Sonnet 4.5, Anthropic is also introducing the Claude Agent SDK. This SDK leverages the same infrastructure that powers Claude Code, enabling developers to construct their own sophisticated AI agents. Additionally, a temporary research preview named “Imagine with Claude” is being released for Max subscribers, showcasing the AI model’s ability to generate software in real-time in response to user requests, without any predetermined functionality or prewritten code.

The current pace of innovation in the AI industry is characterized by rapid production cycles, with companies frequently rolling out flagship models within months. Claude Sonnet 4.5’s launch, less than two months after its predecessor Claude Opus 4.1, exemplifies this trend. Such accelerated development cycles present a challenge for any single company to maintain a significant lead for an extended duration in this intensely competitive field.