

- Big language versions like those developed by OpenAI and Big Tech companies are at risk of being commoditized in 2025, according to executives at leading AI labs.
- Last week, Chinese AI company DeepSeek released R1, a logic type that promises to get better and more cost-effective than OpenAI’s o1 model.
- Tech companies are also promoting the transition from LLMs to so-called “agentic” methods that can integrate these designs and carry out duties on your behalf.
Last week, Chinese AI company DeepSeek released R1, an open-source logic type that claims to competitor OpenAI’s o1 concept on both cost and effectiveness. Open-source program refers to application whose source code is made freely available online for changes and reproduction.
This month, growing recognition of DeepSeek’s new design led to a serious fall in stocks of as investors feared a potential decline in saving on the strong graphics processing units needed to train and manage sophisticated AI workloads.
Nvidia , lost close to$ 600 billion in market capitalization on Monday — , the biggest single-day drop for any company in U. S. history.
In the meantime, managers and researchers at world-class AI laboratories are all promoting a shift from big language versions to so-called” AI brokers” that you carry out actions on your behalf.
LLMs are the basic technology for the conceptual AI apps of today. However, experts believe a drive toward agentic AI systems — , which incorporate LLM technology — , this year will degrade the value of these types.
Commoditization of LLMs
LLMs are expected to become more expensive in the near future as the software gets more sophisticated and the costs associated with education and running them continue to decline.
LLMs will become more knowledgeable systems linked to the agency’s own databases, according to Thomas Wolf, co-founder and chief technology officer of Hugging Face.
” I think people are moving from this madness around the design, understanding that, owing to open-source… a lot of these designs are going to be free and easily available”, Wolf told CNBC. A well-known script collection for open-source AI initiatives is Hugging Face.

” In the internet revolution, we’re moving from building sites as the main business to truly building internet-native companies — therefore, the Airbnb of AI, the Stripe of AI”, he added. ” They are not about the design. They are about the program and how to make the task-specific model important.
Some software CEOs see the release of DeepSeek’s R1 model as another sign that LLMs are becoming more and more expensive.
R1 includes some important characteristics that set it apart from other available designs. For example, the model incorporates a “mixed precision” framework that uses a combination of full-precision 32-bit floating point ( FP32 ) numbers and low-precision 8-bit floating point ( FP8 ) numbers.
The latter can be less precise, but it is quicker to practice. But, DeepSeek switches to FP32 for some jobs where a higher degree of accuracy is required in contrast to concentrate on one or the other for most estimates and uses FP8 for most calculations.
” As AI gets more successful and available, we will see its use rise, turning it into a product we simply can’t get much of”, CEO Satya Nadella on social media platform X, in an apparent reference to DeepSeek.
However, Matt Calkins, CEO of U. S. technology firm Appian, told CNBC that DeepSeek’s success just shows that AI models are going to be more of a commodity in the future.
” In my opinion, we’re going to notice a commoditization of AI. Some companies may reach competitive AI, and a lack of differentiation may be bad for high-spending first-movers”, Calkins said via email.
What are Artificial agencies?
In the last two years, much of the focus has been on LLM-powered products, from relational AI ai like OpenAI’s ChatGPT to resources that can make images and video content.
LLM-based conceptual programs like ChatGPT rely on people to hold their breath while AI agents are made to be more action-oriented.
If you wanted to book a doctor’s appointment, for example, you could ask ChatGPT for information on your nearest GP methods — related to how you’d query the web using Google’s search engine.
Tech executives anticipate that the next generation of AI may include agents who can find out where your nearest GP is, offer applicable times for your physician, and schedule an appointment all directly within a single platform. This is anticipated to change how we communicate with the web.
” Over the years since the internet was invented, the standard interface of pages and entrees hasn’t really changed very much”, Bobby Blumofe, chief technology officer of Akamai, told CNBC recently.
We’re sort of living in the tyranny of having to remember how to navigate web pages and menus while being tethered to a screen.
The shift from models to agents
The CEO of Mistral, Arthur Mensch, stated last week that he believes the “focus should shift to systems” that combines both language models and contextual business data.
” That’s that’s where this is shifting”, Mensch told CNBC’s Arjun Kharpal at the World Economic Forum’s annual gathering in Davos, Switzerland.

Additionally, it implies that the industry adopting it will incorporate its expertise into those systems, which is where it will be required to actually increase the value of AI in those sectors.
Mistral, which is backed by , has swiftly become one of Europe’s buzziest AI firms. Founded only two years ago, it was in June 2024.
Agents are already gaining a lot of popularity. Last week, . It’s trained to interact with” the buttons, menus, and text fields that people use daily” on the web, according to OpenAI, calling it” an agent that can go to the web to perform tasks for you.”
Microsoft, which has invested billions of dollars in OpenAI, in November.
Anthropic, an , -backed AI startup founded by ex-OpenA I research executives, last year introduced” computer use”, which is , a capability that allows its AI agents to use computers like humans to complete complex tasks.
Dario Amodei, CEO of Anthropic, told CNBC’s Andrew Ross Sorkin in Davos last week that Computer Use should be thought of as a precursor to a “broader vision” that integrates these agents with commonly used applications.
” We are planning this year in 2025 to build something that we call the’ virtual collaborator,'” Amodei said, describing it as” an agent that operates on your computer”.

” You can give it tasks. You can say, write this feature for this app, and it’ll write some code, it’ll compile the code, it’ll check the code, it’ll talk to its coworkers on Slack, or on Google Docs, or on some other platform”, Anthropic’s chief explained.
Over the course of the next year, anthropological plans to gradually build out pieces of this vision while keeping it” safe and controllable” are also being made, Amodei added.
PaulO’Sullivan, Salesforce’s chief technology officer for U. K. and Ireland, predicts verbal communication will become more important in the new world of AI agents. In September, Salesforce released its own Agentforce system before Microsoft, OpenAI, or Anthropic, which was arguably its early entry in the agentic AI game.
” We will converse more”, he told CNBC in a recent interview. ” The screens will get smaller, our eyes will get worse. However, they’ll likely be replaced by portable devices that are, in any case, more powerful than laptops.