Claude 3.7 Sonnet by Anthropic
Anthropic has introduced Claude 3.7 Sonnet, a hybrid AI model that offers both real-time and reasoned responses. Users can choose the model's thinking time. It will be available to everyone, but only premium subscribers will access advanced reasoning functions.
Anthropic has launched a new AI model called Claude 3.7 Sonnet. This model is described as the first "hybrid AI reasoning model." Claude 3.7 Sonnet can provide both real-time answers and more considered responses, allowing users to choose whether to activate its reasoning capabilities. Users can decide how long the AI should "think" before providing an answer. This represents Anthropic's effort to simplify the user experience with its AI products, avoiding complexities in model selection. Claude 3.7 Sonnet will be available to all users and developers, but only subscribers to premium plans will have access to the reasoning features. Free users will utilize a standard version that, according to Anthropic, outperforms the previous model, Claude 3.5 Sonnet.
The cost to use Claude 3.7 Sonnet is $3 per million input tokens and $15 per million output tokens. This makes the model more expensive compared to some alternatives like OpenAI's o3-mini and DeepSeek's R1, which are pure reasoning models. However, Claude 3.7 Sonnet offers a hybrid approach that combines reasoning capabilities with quick responses. Anthropic developed this model to address the limitations of traditional methods, making the response process more accurate by breaking problems down into smaller steps.
In the future, Anthropic aims to enhance Claude's capabilities so it can autonomously decide how long to dedicate to a question, without users needing to select options in advance. The model also features a visible internal planning phase for users to observe Claude's thinking process. In practical tests, Claude 3.7 Sonnet achieved 62.3% accuracy in complex coding problems, outperforming other models like o3-mini. Additionally, the model has reduced refusal rates by 45% compared to its predecessor, showing greater ability to distinguish between harmful and benign requests. Alongside Claude 3.7 Sonnet, Anthropic is also launching a coding tool called Claude Code, which allows developers to perform specific tasks through the terminal.
Claude Code provides an easy way to analyze coding projects and make modifications using natural language commands. This tool will initially be available to a limited number of users on a first-come, first-served basis. Anthropic launched Claude 3.7 Sonnet at a time when many companies are rapidly releasing new AI models, marking a shift for the company, which has historically taken a more cautious, methodical approach.
