DeepSeek et AI Industry: High-Performance Open Source Model Disrupts Market | Llm training dataset | Llm vs generative ai | Large language models examples | Turtles AI

DeepSeek et AI Industry: High-Performance Open Source Model Disrupts Market
DeepSeek Challenges AI Giants With a More Accessible, Efficient Open Source Model
Isabella V28 January 2025

 

DeepSeek, with its R1 open source AI model, has shaken up the tech market with its surprisingly low training costs and performance comparable to industry leaders. Former Intel CEO Pat Gelsinger adopts it for his startup Gloo, predicting its incredible impact.

Key points:

  • DeepSeek introduces R1: high performance and low cost.
  • Gelsinger chooses R1 for his Kallm project, abandoning OpenAI.
  • R1 demonstrates the power of open source against the closed models of OpenAI and Anthropic.
  • Skepticism in the industry: accusations of falsified numbers and geopolitical concerns.

DeepSeek’s innovation has caused a stir in the tech world by posing a real challenge to the AI giants with its R1 model, which combines high performance and greatly reduced training costs. Using a data center of 2,000 Nvidia H800 GPUs, DeepSeek completed training its model in just two months at an estimated cost of $5.5 million. This figure, significantly below industry standards, drew both praise and criticism. Pat Gelsinger, former Intel CEO and currently head of the startup Gloo, was among the first to publicly support the technology, saying he chose R1 for his platform’s new AI service, called Kallm. This choice marks a clear departure from the use of OpenAI’s proprietary models, which are considered more expensive and less flexible.

According to Gelsinger, R1 represents a turning point for the industry, demonstrating that excellent results can be achieved by focusing on design ingenuity rather than indiscriminate increases in hardware resources. He pointed out that the open source nature of DeepSeek not only lowers the economic barriers to AI adoption, but also stimulates a more transparent and collaborative ecosystem than the closed models of OpenAI and Anthropic. This vision, according to Gelsinger, could usher in a new era in which quality AI becomes accessible and ubiquitous, influencing not only advanced technology sectors but also everyday devices from wearables to home appliances.

Despite the hype, DeepSeek’s rise has not been without controversy. Some industry experts have questioned the authenticity of the claimed numbers, suggesting that the actual costs may have been higher or that R1’s performance was exaggerated compared to rival models. In addition, the company’s Chinese provenance raised questions about data security and possible censorship. Gelsinger, however, dismissed these concerns, saying that the real lesson of DeepSeek lies in demonstrating to the Western world the advantages of an open source approach, regardless of geographic origin.

The economic and technological implications of this innovation are profound. If confirmed, DeepSeek’s findings could destabilize industry leaders such as OpenAI, forcing them to rethink their business models and development strategies. Not everyone, however, shares Gelsinger’s enthusiasm: many believe that upcoming developments, such as OpenAI’s o3 model, could bring balance back to the market, reasserting the superiority of more expensive and advanced infrastructure.

Gelsinger, as an experienced engineer, remains optimistic: for him, AI must become more widespread, but more importantly, better. DeepSeek’s success, he said, is not just a matter of cost, but of accessibility and quality.

This is an era of great change, and open source is proving that human ingenuity can compete with any budget.