Anthropic launches Claude Sonnet 4.5, its best AI model for coding

On Monday, Anthropic introduced its latest innovation, the Claude Sonnet 4.5, which promises to deliver exceptional performance in coding tasks. This new frontier model is designed to create not just prototypes, but fully functional, production-ready applications, marking a significant advancement in reliability compared to its predecessors. Developers can access Claude Sonnet 4.5 through the Claude API and the Claude chatbot, with pricing unchanged from the previous version at $3 per million input tokens and $15 per million output tokens. This pricing structure translates to approximately 750,000 words, which is more than the entirety of J.R.R. Tolkien's Lord of the Rings trilogy. In the past year, Anthropic's AI models have garnered a strong following among developers and enterprises, attributed largely to their impressive capabilities in software engineering. Notably, companies like Apple and Meta reportedly utilize Claude AI models for internal applications. Anthropic has also built a robust business selling API access for various AI coding applications, including Cursor, Windsurf, and Replit. However, the competitive landscape is heating up. OpenAI's GPT-5 has recently entered the scene, surpassing Claude models on several coding benchmarks. Despite this, Anthropic asserts that the Claude Sonnet 4.5 shines with industry-leading performance on numerous coding evaluations, including the SWE-Bench Verified benchmark. David Hershey, an AI researcher at Anthropic, shared insights with TechCrunch, emphasizing that the true capabilities of Claude Sonnet 4.5 extend beyond mere benchmark results. During early trials with enterprise clients, he observed the model autonomously coding for as long as 30 hours, during which it not only developed an application but also managed database services, registered domain names, and conducted a SOC 2 audit to ensure security. Cursor CEO Micheal Truell touted Claude Sonnet 4.5 as representing the pinnacle of coding performance, especially for long-term projects. Jeff Wang, CEO of Windsurf, echoed this sentiment, noting that this model signifies a new era in coding technology. Anthropic also claims that Claude Sonnet 4.5 is its most aligned AI model to date, showcasing reduced tendencies for sycophancy and deception compared to earlier versions. Furthermore, the company has enhanced Claude’s resilience against prompt injection attacks. Alongside the launch of Claude Sonnet 4.5, Anthropic is introducing the Claude Agent SDK, which serves as the backbone for Claude Code and assists developers in creating their own agents. Additionally, a temporary research preview called 'Imagine with Claude' will be available for Max subscribers, showcasing the AI model's ability to generate software dynamically in real time without prewritten code. As the AI industry continues its rapid evolution, with companies frequently releasing flagship models, Claude Sonnet 4.5 arrives just two months after the introduction of Anthropic's previous model, Claude Opus 4.1. This fast-paced competition makes it challenging for any single company to maintain a significant lead for an extended period.