Anthropic, an AI company, released a new version of its Claude software called Claude Sonnet 4.5 on September 29, 2025. The company built this version specifically to be better at two things: writing computer code and running automated systems that can solve multi-step problems without human help.

Think of autonomous agents like a virtual assistant that can break a large task into smaller pieces, use different tools to complete each piece, and keep track of everything as it goes. Building reliable systems that work this way is difficult — they often get confused or make mistakes partway through a task. Anthropic designed Sonnet 4.5 to be more stable at this kind of work.

The company is selling this model directly to businesses and companies, not to individual consumers. Reuters reported that this is part of Anthropic's overall business strategy. Coding help and automated agents are valuable services that many companies want to pay for.

Anthropic's previous version, Claude 3.5 Sonnet, was made free for regular people to use on the web and on phones. That helped many programmers learn the software. This new version keeps the same name, but Anthropic is selling it to companies instead, charging by how much they use it.

Right now, many AI companies are competing to say they have the best coding model. OpenAI, Google, and others have all released models with strong performance at writing code. When Anthropic says Sonnet 4.5 is the best at coding, the natural question is: best compared to what, and how do we measure that? Anthropic has not released full test results from independent outside reviewers at the same time as the announcement — something companies will want to see before making big commitments.

The more important test will come in the real world. Programmers building actual automated systems will quickly find out whether Sonnet 4.5 really works better than what they have now. The claims on marketing websites matter less than what actually happens when people try to use it.

Anthropichas been releasing new versions of Claude faster than many expected. The gap between the last version and this one is about fifteen months, during which the entire AI industry shifted multiple times. Whether Sonnet 4.5 stays the leader in coding and automation through the next wave of competitor releases will depend on what those companies build next.