Anthropic to introduce AI with adjustable depth of thinking

Anthropic is in the final stages of preparing to launch a new artificial intelligence model, which could appear in the coming weeks. According to The Information, the development represents a revolutionary hybrid approach combining deep analysis capabilities with quick responses.

A key feature of the new model will be an innovative “sliding scale” system that will allow developers to control the balance between performance and computational cost. This is particularly important when using deep thinking functions that require significant computational resources.

According to available data, Anthropic’s new development is already showing impressive results, surpassing OpenAI’s o3-mini-high model in several programming tasks. The model’s strengths are particularly evident in analyzing large codebases and solving business-oriented tasks.

Anthropic CEO Dario Amodei commented on the company’s AI development philosophy in a recent TechCrunch interview: “We’re generally focused on creating our own approach to reasoning models that are better differentiated. We’re somewhat puzzled by the idea that there are regular models and reasoning models, and that they’re somehow different from each other.”

This approach reflects Anthropic’s desire to overcome the traditional division between models optimized for speed and models oriented toward deep analysis. The new development could be an important step in developing more flexible and universal artificial intelligence systems capable of adapting to various tasks and user requirements.