Anthropic, an AI safety and research company that’s working to build reliable, interpretable, and steerable AI systems, has announced the launch of Claude Sonnet 4.5, a cutting-edge AI agent capable of autonomously managing complex, multistep coding tasks for up to 30 hours straight.
Branded as “the best coding model in the world,” the company describes the tool as the strongest model for building complex agents, noting that it is the best model at using computers.
Announcing the launch via a blog post, Anthropic wrote,
“Claude Sonnet 4.5 is the best coding model in the world. It’s the strongest model for building complex agents. It’s the best model for using computers. And it shows substantial gains in reasoning and math. Code is everywhere. It runs every application, spreadsheet, and software tool you use. Being able to use those tools and reason through hard problems is how modern work gets done.
“Claude Sonnet 4.5 makes this possible. We’re releasing it along with a set of major upgrades to our products. In Claude Code, we’ve added checkpoints, one of our most requested features, that save your progress and allow you to roll back instantly to a previous state. We’ve refreshed the terminal interface and shipped a native VS Code extension. We’ve added a new context editing feature and memory tool to the Claude API that lets agents run even longer and handle even greater complexity.”
The new model features enhanced reasoning and mathematical capabilities, making it highly valuable for specialized industries such as cybersecurity, finance, and research. Experts in finance, law, medicine, and STEM found Sonnet 4.5 shows dramatically better domain-specific knowledge and reasoning compared to older models, including Opus 4.1.
Claude Sonnet 4.5’s debut comes at a time when OpenAI is also rolling out parental controls and shopping features for ChatGPT, intensifying competition in the rapidly evolving AI landscape.
The recently released model sets a new benchmark for AI-driven coding, with industry test results, including the SWE-bench Verified benchmark, validating its superior software development skills. According to Jared Kaplan, Anthropic’s co-founder and chief science officer, the model’s intelligence makes it feel more like a colleague than a tool, improving collaboration between developers and AI systems.
“People are just noticing with this model, because it’s smarter and more of a colleague, that it’s kind of fun to work with it when encountering problems and fixing them,” Kaplan told CNBC.
The model generates higher-quality code, identifies potential improvements, and follows instructions more reliably than previous versions. It also significantly extends autonomous run time from seven hours with Claude Opus 4 to 30 hours, making it ideal for long-term, multistep projects such as software testing and cybersecurity operations.
Mike Krieger, Anthropic’s chief product officer, confirmed that Claude Sonnet 4.5 will be the default model for most users, recommended for “basically every use case.” However, users will retain flexibility, as paid subscribers can still access Claude Opus or earlier Sonnet models if needed.
Krieger emphasized that while Claude Sonnet 4.5 is smaller than Claude Opus 4.1, it is “smarter in almost every single way. We have found it, and our customers are finding it, very useful for real, actual work,” he said.
Anthropic has focused heavily on improving safety, addressing concerns about AI misuse and malicious attacks. The company says Claude Sonnet 4.5 is now far more resistant to prompt injection attacks, where hackers attempt to manipulate AI into leaking sensitive information or performing harmful actions.
The model also exhibits fewer problematic behaviors, including deception, power-seeking tendencies, and sycophancy—a behavior where the AI simply tells users what they want to hear. Kaplan called these improvements “the biggest jump in safety we’ve seen in the last year to year and a half.”
Claude Sonnet 4.5’s launch coincides with a global surge in Claude adoption, with 80% of usage now coming from outside the U.S. Adoption rates in countries like South Korea, Australia, and Singapore have already surpassed U.S. levels on a per-person basis.
The launch of this model comes as Anthropic announced plans to triple its international workforce and expand its applied AI team fivefold in 2025. The company has begun hiring country leads in India, Singapore, Australia, New Zealand, and South Korea, with further expansion planned across Europe, including the UK, Nordics, Southern Europe, and the DACH region.
Anthropic’s rapid rise is reshaping the global enterprise AI market. The company has achieved a $5 billion revenue run-rate, a dramatic leap from $87 million at the start of 2024. This growth positions Anthropic as a formidable rival to OpenAI, Microsoft, and Google, with a distinct strategy: delivering a pure-play AI platform rather than integrating its technology as a secondary feature within legacy systems.
Looking forward
Claude Sonnet 4.5 release underscores Anthropic’s ambition to outpace competitors like OpenAI and Google in the race to build advanced programming agents capable of solving real-world business and cybersecurity challenges.
Notably, the company has hinted at one or two more model launches before the end of 2025, including a likely upgrade to the Opus series. With Claude Sonnet 4.5, Anthropic has not only raised the bar for AI-powered coding agents, but also signaled its intent to dominate the next era of autonomous, enterprise-grade AI solutions.
