Claude 3.7 Sonnet It is the new artificial intelligence model developed by Anthropic And it represents a significant evolution with respect to its predecessors. Its distinctive feature, which makes it the first to of this type, is the “Hybrid” reasoninga capacity that allows him to alternate between quick responses and more in -depth analyzes according to the context. This makes it particularly versatile, improving its performance in areas such as mathematics, programming and resolution of complex problems. Travelers who access the model via Bee (Application Programming Interface) can customize the duration of its processing to find a balance between speed and quality of responses. As Anthropic has made known, Claude 3.7 Sonnet “It is now available on all Claude plans, including Free, Pro, Team and Enterprise, as well as on Anthropic API, Amazon Bedrock and Vertex Ai Di Google Cloud». It must be said, however, that the Method of extended thinking – the one in which the model reflects on itself before responding – is not available for free accounts.
In addition to the improvements in reasoning, it introduces Claude Codean advanced tool for developers who facilitate writing and code management. The first tests show that this model has excellent performance in the coding and management of complex working flows. Anthropic tested Claude 3.7 Sonnet even in the video game Red Pokémonwhere he managed to progress significantly compared to previous versions. A further signal that the model has improved a lot compared to its predecessor.
How Claude 3.7 Sonnet works, the artificial intelligence of Anthropic
Unlike the previous models, which tended to separate the rapid reasoning from that more in -depththe new model Claude 3.7 Sonnet integrates both modalities in a single architecture. This means that it can answer instantly or take more time to process more articulated answers, offering users unprecedented on the quality and depth of its responses.
Anthropic was inspired by the functioning of the human brain to implement this important novelty. In the official press release with which he announced Claude 3.7 Sonnet in fact said:
Just as human beings use a single brain both for quick responses and for in -depth reflections, we believe that reasoning must be an integrated ability of border models rather than a completely separate model. This unified approach also creates a more fluid experience for users.
A key aspect of Claude 3.7 Sonnet is the possibility of Manage the so -called “Budget for thought” through API. Basically, developers can set up a maximum token limit to be used for reasoning, allowing a balance between the processing time and the cost of the operation. Ah … for the record, tokens are the fundamental units of processing in language -based artificial intelligence models, such as Claude in fact. A token can correspond to a whole word, a part of speech or even a single character, depending on the language and context.
On the front of the programmingClaude 3.7 Sonnet has achieved significant improvements compared to his predecessors. His ability to understand and generate code has been refined thanks to greater attention to real use cases, rather than to academic competitions of mathematics and IT. In the tests conducted by companies such as Cursor, Cognition And Canvathe model has proven to excel in the management of complex code bases, in the modification of files, in full-stack development and in the optimization of the code. According to the assessments, the code produced by Claude is more accurate and requires less corrections than that generated by the previous models.
There safety It is another aspect that Anthropic has worked significantly. Claude 3.7 Sonnet was trained to better distinguish between legitimate and potentially harmful requests, reducing the number of unjustified waste of 45% compared to the previous version. In addition, it was designed to resist cyber attacks, such as those of prompt injectionin which cybercriminals try to deceive the artificial intelligence models by inducing them to provide unwanted answers by the user through malicious instructions which, without the knowledge of the latter, are given to the model.
The introduction of the Claude Code
A revolutionary element introduced with this version is Claude Codea new tool designed to assist developers in an even more advanced way. Claude Code does not limit himself to suggesting code fragments, but is able to search and analyze repository, edit files, perform tests and load changes directly on platforms such as Github. This makes it a real Digital assistant for programmerscapable of drastically reducing development times and improving work efficiency. If you want to deepen the characteristics of Claude Code, you can do it with the following video (in English) made by Anthropic.
https://www.youtube.com/watch?v=ajpk3Yttkz4
Claude 3.7 Sonnet struggling with video games
In addition to the practical applications in the programming and resolution of complex problems, Anthropic tested Claude 3.7 Sonnet even in a more unusual context: the Red Pokémon video game. Unlike previous versions, which could not even overcome the initial stages of the video game, this new version managed to defeat three capaalestra and collect their medalsdemonstrating advanced planning and adaptation skills. Although this test may seem only a curious experiment, it actually provides precious indications on the potential that the model has in facing complex tasks that require memory, strategic planning and progressive learning.
