Anthropic’s Claude 3.7 Sonnet takes aim at OpenAI and DeepSeek in AI’s next big battle

Anthropical just hit a warning Open, Depth and by launching the entire AI industry Claud 3.7 Sonneta model that makes an unprecedented control for how much an AI spends users “think“Before the answer to the answer. Along with the debut Claud EncodingA team line AI coding agent, an aggressive charges of anthropy to an enterprise to the enterprise to the AI ​​Market – a boost that can automate businesses and automate their work.

The stakes could not be higher. Last month DeepSeek, Tech World with AI model with AI model AI Price fractionSending Nvidia stock 17% and increase the alarms about the AI ​​management of America. Now the anthopic AI bets on accurate control of justification – it will not only save raw speed or cost – an edge.

“We just believe that the justification is not a major part of the EU, but not a reasonable thing,” Dianne Penn, VentureBeat-in an interview with Dianne Penn. “Like people, AI must manage both fast answers and complex thinking. ‘When is it?’ It should be answered immediately for a simple question. But for complicated tasks – love to plan a two-week-old Italian trip when placing the gluten-free diet needs – a wider processing time. “

“We do not see justification, planning and self-correction as separate opportunities.” “Thus, this is the way this philosophical difference is … Ideally, the model itself requires more to recognize a problem than demanding more intensive thinking modes of thinking and regulation.”

Assessment data supports anthrop’s ambitious vision. In Extended Thinking mode, Claude 3.7 Sonnet achievements 78.2% accuracy In graduating level justification tasks, Openai has difficulty in the latest models and superior DeepSeek-R1.

However, more detected measurements come from real world applications. Model scores 81.2% is the vehicle used in retail focus and indicates marked developments Instructions – the following (93.2%) – places where competitors are fighting or not.

While DeepSseek and Openai Leads Traditional Mathematics CriteriaClaude 3.7 demonstrates a single model, a single model can effectively pass between fast answers and deep analysis, eliminates the need for enterprises to protect separate AI systems for different types of tasks.

How can anthrop a hybrid EU change the computation of the enterprise

Issue time is important. Deepseek’s emergence was sent last month Shocking waves By means of silicon valley, you can achieve an advanced AI substantiation Fewer calculation power than remote more than previously thought. This protested against fundamental assumptions related to AI development costs and infrastructure requirements. Foundation of NVIDIA when DeepSEEK published the results 17% fell In one day, investors ask if suddenly expensive plugs are important for the developed AI.

Share for businesses could not be higher. Companies Spend millions To integrate the EU operations, the bet on which approach will prevail. Anthropik’s hybrid model offers an attractive medium way: Ability to delicate tune performance based on the task that comes to a task immediately. The system protects anthropic Previous price With a million dollar troops for a million dollar tokens and a million dollars worth a million dollars, even with additional justification features.

“Our customers are trying to achieve results for customers,” Michael Gerstenhaber, the head of the anthopic platform. “Using the same model and allow someone to ask for the same model in different ways Thompson Reuters Allows our coding partners to conduct legal research Cursor or Entrusted To develop applications and meet these goals. “

The hybrid approach of anthrop is both a technical evolution and a strategic gambit. While keeping Openai Separate models for different opportunities and DeepSeek draws attention cost efficiencyAnthropic is followed by unit systems that can manage regular tasks and complex thinking. It is a philosophy that can eliminate the needs of the enterprises to place AI and the need to juggle the models of many specialized models.

Meet Claude code: AI’s new developer assistant

Anthropic is still announced today Claud EncodingThe developers are a command line that allows you to display complex engineers directly to the AI. The system requires a person’s approval before making code changes, aiming to develop an increasing industry in charge of the growing industry.

“You really need to accept changes yet. You are a reviewer with your hands [the] Wheel, “Penns noted.” There is a kind of checklist you should essentially accept certain actions to make certain actions. “

Ads are in intense competition in the development of AI. Stanford researchers Recently created Open Source Justification Model Less than $ 50, Microsoft is integrated only Openai’s O3-Modi Model Has become azure. Deepseek’s success, as well as new approaches to the development of the AI, with some companies that can further reduce costs, new approaches were spread.

From Pokémon to Enterprise: Testing the new intelligence of AI

Penn described the sharp progress in the AI ​​capabilities, which is an unexpected example: “I asked Claude’s different versions of Claude to play … This version showed it all the way Vermilion cityMore than one Pokémon seized and even rose to the level. There is the right pokemon to fight opponents. “

“I think we will see us to see if we continue to focus on things such as innovative and reasoning, and we continue to direct them to things like a dynamic justification.”

The real test of the anthropy approach will come from the admission of the enterprise. Pokémon may seem trivial when playing, it demonstrates the EU that can make both regular operations and complex strategic decisions without the transition between specialized models. Previous versions of the clod could not go outside of a game starting city. The latest version builds strategies, manages sources and makes tactical decisions – the opportunities reflecting the complexity of real world business problems.

For enterprise customers, it can express many AI systems for different tasks and express the difference between placing a solution to a solitary solution. The next few months will reveal the fact that Anthrop is changed the enterprise market in the formation of the Ai’s justification of the EU or only another experience in the industry’s rapid evolution.