OpenAI vs. Meta: Hidden Chain of Thought vs. Open Source
OpenAI’s o1 model is an improvement over GPT 4o. However, will OAI’s decision to hide the chain of thought allow Meta’s Llama models to capture market share given Llama’s open source approach?
OpenAI dropped a preview of its next frontier model - o1 - today. o1 is a significant improvement over GPT 4o along Math and Science disciplines (below).
Perhaps the most interesting element of the o1 preview is that OpenAI decided to hide the chain of thought from users (and competitors).
A hidden chain of thought is 180 degrees different from META’s open source approach. I can imagine Google doing the same with Gemini and Anthropic hiding Claude’s chain of thought.
The $125 billion question is: “Will Meta’s Llama family of LLMs take market share as a result?”