Humanity launches Claude 4.5 to promote better capabilities and target business customers

Humanity announced the Claude 4.5 AI model on Monday, saying the latest version could code longer, uninterrupted stretches, and handle financial and scientific tasks better as startups push enterprise AI deeper.

Supported by Alphabet and Amazon.com, AI startups are rivals for building models that can reliably operate software and complete multi-step tasks, and are keys for AI agents that can perform tasks on behalf of humans.

The Sonnet 4.5 model created web apps from scratch in internal testing, with one customer autonomously having an AI chatbot code for 30 hours.

He said humanity is targeting power users and business customers rather than chasing viral consumer moments.

Claude 4.5 is stronger in financial and scientific reasoning, better computer use, and in previous models it has earned around 60% in benchmarks that test the dexterity of the operating system and about 40%, the company said.

“If you're not a coder, you just use a computer to look at the model in a person's way, and it's much more visceral,” said Jared Kaplan, Chief Science Officer.

Microsoft said on Monday that it will add a separate new Microsoft 365 Copilot feature with humanity models, including “agent mode” for Excel and Word and “office agent” for Copilot Chat.

Last week, Microsoft said it would bring the human model to Microsoft 365 Copilot, diversifying beyond its longtime partner Openai.

Founded by former Openai executives, humanity positioned Claude for use in the Guardrails in the workplace. The company sells Claude coding and data analytics skills to regulated industries and teams who want their models to work with multiple software tools.

Krieger said the company's focus is not on short demos but on sustainable and reliable performance over long tasks.