As AI technology continues to evolve at a rapid pace, executives at leading AI labs are predicting that large language models (LLMs), such as those developed by OpenAI and major tech firms, may face commoditization in 2025, CNBC reports.
With the introduction of innovative systems like DeepSeek’s R1 reasoning model and the growing focus on AI agents, the value of traditional language models may decrease as more versatile and cost-effective alternatives emerge.
DeepSeek, a Chinese AI firm, recently unveiled its R1 model, claiming that it outperforms OpenAI’s o1 model both in terms of cost and performance. R1 also introduces a “mixed precision” framework that combines both full-precision and low-precision calculations, offering greater efficiency. As a result, the model is gaining attention for its ability to provide high-level performance while maintaining affordability. This development is expected to further push the commoditization of LLMs, as more open-source models like DeepSeek’s become readily available.
The rise of open-source AI and the decreasing costs associated with training advanced models have led many experts, including Thomas Wolf of Hugging Face, to suggest that LLMs will eventually become a commodity. Hugging Face, a major player in the open-source AI space, points out that, as LLM technology becomes increasingly integrated into intelligent systems, models will be freely available and accessible to anyone. This shift will likely be analogous to the internet revolution, where the focus moved from building websites to building internet-native companies that leverage the underlying technology for specific tasks.
Moreover, DeepSeek’s R1 model is just one example of how the landscape is shifting. Executives are now pointing to the emergence of “agentic” AI systems as the next frontier in AI development. These AI agents go beyond traditional LLMs by performing tasks on behalf of users, making them action-oriented rather than simply responding to queries. For example, instead of asking an AI to provide information on available doctor’s appointments, a future AI agent could autonomously book the appointment for you, all within a single platform.
The transition from LLMs to AI agents is being driven by companies such as OpenAI, Microsoft, and Anthropic. OpenAI has already launched “Operator,” an AI agent capable of performing tasks like interacting with web buttons and text fields. Similarly, Anthropic’s “computer use” functionality allows AI agents to operate computers, handling complex tasks like writing and debugging code, and even collaborating via platforms like Slack or Google Docs.
This shift to AI agents is expected to further diminish the role of traditional LLMs, as the focus moves from the model itself to the broader systems that integrate these models with real-world applications. The ability of AI agents to handle practical tasks in an autonomous manner offers users more seamless and efficient experiences, which could ultimately make the underlying language models less central to the value proposition of AI products.
In addition to shifting the value of LLMs, the rise of AI agents could have broader implications for industries and the workforce. As AI becomes more capable of taking on complex tasks, the need for human intervention may decrease in some areas, leading to new business models and potentially altering how businesses operate.
Despite the growing interest in AI agents, many experts believe that LLMs will continue to play an important role in the development of future AI systems. However, the growing competition from both open-source models and agent-based systems suggests that the market for LLMs may become more crowded and less profitable for large tech firms, with companies like DeepSeek and others pushing for more cost-effective and efficient solutions.
