Antarbur launches a new model of artificial intelligence “thinking” as long as you want

Antarbur launches a new model for Frontier AI called Claude 3.7 Sonnet, which designed the company “thinking” in questions as long as users want it.

Claude 3.7 Sonnet calls the first “AI Hybrid AI” model, because it is one model that can give both actual and more “thoughtful” answers to answers to questions. Users can choose whether the “thinking” capabilities of the artificial intelligence model will be activated, which causes Claude 3.7 Sonnet to “thinking” for a short or long period of time.

The model is the broadest effort of anthropology to simplify the user experience about its AI products. Most AI Chatbots today contain a hard model that forces users to choose from many different options that differ in cost and ability. Laborators like Antarbur prefer not to have to think about it – ideally, one of the models does all work.

Claude 3.7 Sonnet said to all users and developers on Monday, and Anthropor said, but only people who pay the distinctive Chatbot plans from Antarbur will be able to reach the features of thinking about the model. Claude users will get the unusual standard version of Claude 3.7 Sonnet, which is surpassed Claude 3.5 Sonata. (Yes, the company exceeded a number.)

Claude 3.7 Sonnet costs $ 3 per million input codes (which means that you can enter approximately 750,000 words, words more than the entire “Lord of the Rings” series, to Claude for $ 3) and $ 15 per million output symbols. This makes it more expensive than Openai’s O3-MINI ($ 1.10 per million input codes/$ 4.40 per million output symbols) and Deepseek’s R1 (55 cents per million input codes/$ 2.19 per million produced codes), but consider that O3-MINI And R1 are strict-not hybrid thinking models like Claude 3.7 Sonnet.

Claude 3.7 Sonnet is the first model of AI of Hiop Several artificial intelligence laboratories have turned into traditional methods to improve AI’s performance.

Use thinking models such as O3-Mini, R1, Gemini 2.0 Flash Thinking, Xai’s Grok 3 (Think) use more time and computing power before answering questions. Models are divided into smaller steps, which tend to improve the accuracy of the final answer. Thinking models do not think or think like a person, necessarily, but their process is designed after the opponent.

Ultimately, Antarubor would like to know the knowledge of the time you should “think” about the questions on her own, without the need to choose users in advance, Diane Benspers, Diane Ben, Techcrunch said in an interview.

“Like how humans do not have separate brains of the questions that can be answered immediately in exchange for those that require thinking,” Anthroprobi wrote in A. Blog post Common with Techcrunch, “We consider logic just one of the capabilities that the border model must enjoy, to be smoothly combined with other capabilities, rather than something that must be provided in a separate model.”

Anthropor says it allows Claude 3.7 Sonnet to show the internal planning stage through a “visible scratch plate”. Penn told the Techcrunch users will see the full thinking process for most of the claims, but some parts may be revised for trust and safety purposes.

Anthropor says it improves Claude’s thinking modes of real tasks, such as difficult coding problems or agent tasks. Developers who click on the human applications programming interface can control the “budget” for thinking, rapid trading and the cost of the answer quality.

In one test to measure coding tasks in the real word, Claude 3.7 Swe-Bench was 62.3 %, compared to the Openai’s O3-MINI model that recorded 49.3 %. In another test to measure the ability of the artificial intelligence model to interact with simulating users and an external application programming interface in retail sale, Claude 3.7 Sonite 81.2 %, compared to the Openai model that recorded 73.5 %.

Antholbropi also says that Claude 3.7 Sonite will refuse to answer questions less than their previous models, claiming that the model is able to make a more accurate discrimination between harmful and benign claims. Anthropor says it reduces unnecessary rejection by 45 % compared to CLAUDE 3.5 Sonnet. This comes at a time when Some artificial intelligence laboratories re -think about their approach to restricting AI Chatbot answers.

In addition to Claude 3.7 Sonnet, Antarbur also launches an agent coding tool called Claude Code. After launching the search inspection, the tool allows developers to run specific tasks through Claude directly from their station.

In the experimental offer, anthropron employees showed how Claude code can analyze the coding project with a simple matter such as, “Explain this project structure. Using the regular English in the command line, the developer can adjust the code base.

A spokesman for Antarbur told Techcrunch that Claud Code will be available at the beginning of a limited number of users on the basis of “First Come and First Serving”.

Anthropor Claude release 3.7 Sonit at a time when AI LABS charges new artificial intelligence models at a break. Historically, Antarbur has taken a more systematic approach that focuses on safety. But this time, the company is looking to lead the package.

How long, though, is the question. Openai may be about to launch a model of its own hybrid artificial intelligence; The company’s CEO, Sam Al -Tamman, said that it will arrive in “months.”