- Anthropic released Claude Sonnet 4.5, calling it the top AI coding system.
- The startup says the new model can operate autonomously for up to 30 hours.
- Claude Agent SDK aims to give developers new tools for building customized, context-aware AI agents.
Anthropic released Claude Sonnet 4.5, its latest model, on Monday. The company positioned it as the world's best AI coding system and a leap forward in applied artificial intelligence.
The upgrade arrives just four months after its predecessor, Sonnet 4, underscoring the startup's rapid product cadence in the generative AI arms race.
Anthropic said Sonnet 4.5 delivers state-of-the-art results on SWE-Bench Verified, a standard for evaluating software engineering performance.
The startup also pitched the new model's ability to generate practical business outcomes through autonomous computer use, cybersecurity capabilities, and the creation of production-ready applications and context-aware AI agents.
Anthropic's revenue has surged this year, primarily driven by the coding functionalities of its models and a specific product called Claude Code. The startup pulled away from rivals in AI coding, and Sonnet 4.5 is designed to maintain this lead.
Automated and assisted software coding is one of the most compelling use cases for generative AI so far. That's partly because there are big potential productivity gains and cost savings.
Anthropic noted on Monday that Claude Code is generating more than $500 million in run-rate revenue, with usage growing more than 10X in three months.
The startup said Sonnet 4.5 enhances code reliability, refactoring judgment, and production-readiness. This new model competes against other offerings such as Google's Gemini, OpenAI's GPT-5, and xAI's Grok 4.
The business implications are broad. In cybersecurity, Anthropic said Sonnet 4.5 helps detect and remediate vulnerabilities faster. In financial services, it surpasses Anthropic's Opus 4.1, the company's most advanced reasoning model, in tasks such as research, modeling, and forecasting.
Perhaps most notably, the new model can operate autonomously for up to 30 hours, more than quadrupling the endurance of Opus 4, Anthropic said.
The startup is also expanding its developer ecosystem with new tools that bring Claude Code's building blocks to a wider audience. Developers gain access to virtual machines, memory, and context management.
New developer-focused features include a native VS Code extension, enhanced terminal workflows, and checkpoints that allow engineers to roll back code instantly if their AI-powered projects veer off track.
On the Claude Developer Platform, Anthropic launched a Claude Agent software development kit providing developers with fine-grained tools for building customized, context-aware AI agents.
Sign up for BI's Tech Memo newsletter here. Reach out to me via email at [email protected].
