Ai+ Training

The ability to perform complex reasoning is one of the basic requirements for truly intelligent AI models. A new generation of LLMs, such as the o1 and o3 family of models from OpenAI or DeepSeek’s new models, defied many experts’ expectations by smashing reasoning benchmarks designed with AGI in mind. However, these new models are still traditional LLMs at their core. What makes these models so different? How do they “reason”? What are the best usecases for them, and when is better to use a regular LLM such as GPT 4o or Llama? In this talk we will look beyond the hype, get an intuitive understanding of the way they work, and understand the pros and cons of this new crop of models.