Anthropic yesterday unveiled Claude Sonnet 4.5 , which the company describes as the “ world’s best coding model ,” outperforming GPT-5 and Gemini 2.5 Pro. It’s also the most powerful model for developing complex agents and computing , and Anthropic says it features significant improvements in logic and mathematics compared to Opus 4.1.
Along with Claude Sonnet 4.5, Anthropic added a new terminal interface and checkpoints to Claude Code. Checkpoints act as save points, saving your progress so you can revert to a previous state if necessary.
See also: New Agent Mode: Microsoft introduces 'vibe working' in Excel & Word
Claude apps now support code execution and file creation to create spreadsheets, presentations, and documents directly in chat. Claude for Chrome is also available, with Anthropic allowing some Max users to test the feature in beta. There is also a Claude Agent SDK available for agent development.
According to Anthropic, Claude Sonnet 4.5 features enhanced capabilities and extensive safety training to improve its behavior, resulting in a reduction in submissiveness, deception, power seeking, and the tendency to encourage delusional thoughts.
Claude Sonnet 4.5 is available everywhere and Anthropic recommends users upgrade for a better experience.
Claude Sonnet 4.5: What's behind the new upgrades
The above improvements show that Anthropic’s Claude Sonnet 4.5 is not just another upgrade in a rapidly evolving field; it is a strategic move that is reshaping the competition in the large language model. The company’s report that the new model outperforms both OpenAI’s GPT-5 and Google’s Gemini 2.5 Pro demonstrates Anthropic’s confidence and ambition to establish itself as a strong player in front-line AI.
See also: Nvidia to invest up to $100 billion in OpenAI
What stands out is not only the technical improvements in coding and mathematical logic, but the direction towards tooling AI in real production flows. The new terminal interface and checkpoints in Claude Code are a clear indication that Αnthropic aims not just to demonstrate power, but to increase usability for developers and businesses seeking stability and transparency in use.
Of particular interest is the integration of file creation capabilities – from spreadsheets to presentations – directly into the conversation. This is a step that moves AI from the role of “advisor” or “text assistant” to that of an active partner in content production, connecting Anthropic’s direction with the vision of an “AI office suite” that many companies are pursuing.
The availability of a Claude Agent SDK also paves the way for a third-party application ecosystem, where independent developers can create specialized agents. This strengthens Anthropic's advantage over more closed systems and gives impetus to a more decentralized approach to developing practical AI tools.
See also: DeepSeek-R1-Safe: New AI model with the help of Huawei
At the same time, a systematic emphasis on security, with training to reduce misleading responses, is a key competitive advantage. At a time when regulators are focusing on the implications of large models, Anthropic is attempting to get ahead of the curve by presenting Claude Sonnet 4.5 as a powerful yet more responsible technology.
With these moves, the company shows that it sees AI not only as a means of innovation, but also as an infrastructure for productivity. The big question is whether the market – from businesses to individual users – will recognize the new Claude as the platform to integrate into their daily work, in a space where OpenAI and Google still hold the upper hand.
