Meta’s New AI Model Chameleon to Challenge GPT-4o by OpenAI

MetaversePlanet May 24, 2024

0 1 minute read

Meta is developing a new artificial intelligence model called “Chameleon” to compete with OpenAI’s GPT-4o model. However, Chameleon is still in its early stages.

OpenAI recently announced their new artificial intelligence model, GPT-4o, at an event held in the past few weeks, which has significantly impacted the sector. The new AI model can perceive its environment, prompting OpenAI’s competitors to start similar projects.

In this context, it has been revealed that Meta is also working on a new AI model named “Chameleon.” This model converts various types of data into a set of tokens, which are then integrated to work harmoniously. Chameleon will be known as a multimodal artificial intelligence.

Trained with a massive dataset of 4.4 trillion

Meta's New AI Model Chameleon to Challenge GPT-4o by OpenAI

Chameleon is the result of a comprehensive study. Meta trained this artificial intelligence model with an extensive library containing 4.4 trillion pieces of text, audio, and images. The AI was trained on high-speed GPUs for 5 million hours, utilizing 34 billion parameters in the process. In comparison, OpenAI uses around 1 trillion parameters for its training.

Meta’s new AI model is still in its early stages, meaning it has not yet reached the end user. However, this is expected to change in the near future. It will be interesting to see how Meta’s new AI model will compete with Google’s Gemini and OpenAI’s GPT-4o.