Anthropic has officially launched Claude Sonnet 4.5, its most advanced AI model for coding to date, promising state-of-the-art performance and the ability to build production-ready software applications rather than mere prototypes.
Available via the Claude API and Claude chatbot, Claude Sonnet 4.5 retains the same developer pricing as its predecessor—$3 per million input tokens and $15 per million output tokens. That translates to roughly 750,000 words per million input tokens, keeping it affordable for enterprises and developers alike.
Raising the Bar for AI Coding Performance
Over the past year, Anthropic’s AI models have become a favorite among software developers, enterprises, and coding platforms such as Cursor, Windsurf, and Replit. Tech giants, including Apple and Met,a reportedly use Claude AI models internally. However, the competitive landscape intensified recently when OpenAI’s GPT-5 began outperforming earlier Claude versions on multiple coding benchmarks.
With Claude Sonnet 4.5, Anthropic aims to reclaim its edge. The company reports industry-leading results on benchmarks like SWE-Bench Verified. Yet, according to Anthropic researcher David Hershey, the model’s true capabilities extend beyond test scores.
“In enterprise trials, we watched Claude Sonnet 4.5 autonomously code for up to 30 hours,” Hershey revealed. “It didn’t just write applications—it configured databases, purchased domains, and even ran SOC 2 security audits.”
Industry Leaders Praise the Launch
Tech leaders are already weighing in. Cursor CEO Michael Truell calls Claude Sonnet 4.5 “state-of-the-art for long-horizon coding tasks,” while Windsurf CEO Jeff Wang describes it as “a new generation of coding models.”
The launch also comes with notable enhancements: improved resistance to prompt injection attacks, lower rates of AI sycophancy, and stronger overall alignment to developer needs.
New Tools for Developers
Alongside Claude Sonnet 4.5, Anthropic introduced the Claude Agent SDK, enabling developers to build custom AI agents using the same infrastructure that powers Claude Code. Additionally, a research preview called Imagine with Claude will allow Max subscribers to watch real-time code generation without predefined constraints.
Racing Ahead in AI Innovation
The release follows Claude Opus 4.1 by less than two months, highlighting the fast-paced competition in AI development cycles. With Claude Sonnet 4.5, Anthropic hopes to set a new benchmark for enterprise-grade AI coding solutions.
As rapid innovation continues, one thing is clear: the battle for AI dominance in coding has only just begun.