The Largest AI Models and Their Training Scale

The race to train increasingly massive AI models with billions and even trillions of parameters is a defining characteristic of current AI development, requiring immense computational resources and pushing the boundaries of what's possible.

The top 3

  1. The Three AI Models with the Most Parameters: Kimi K3, with 2.8 trillion parameters, has been announced as the world's largest open-source AI model, while other frontier models like those from OpenAI and Google are also estimated to be in the trillions of parameters, such as a GPT-MoE model cited at 1.8 trillion parameters.
  2. Top Three Companies Investing in AI Supercomputing: Google, Microsoft (in partnership with OpenAI), and NVIDIA are major investors in AI supercomputing infrastructure, with Google aiming for one million TPU chips and NVIDIA planning significant investments in OpenAI's infrastructure.
  3. Three Major Milestones in AI Model Scaling: Key milestones in AI model scaling include the conceptualization of artificial intelligence by Alan Turing in the 1950s, the coining of "artificial intelligence" at the 1956 Dartmouth Conference, and the emergence of models with trillions of parameters in the 2020s, significantly advancing computational power and data processing capabilities.

Sources

Open the full topic