Preeti
Meta has released a new family of AI models called Chameleon, which can handle images and text with ease. These models, comparable to commercial tools like Gemini Pro and GPT-4V, are capable of understanding and generating images and text. Chameleon comes in 7 billion and 34 billion parameter versions and can process combinations of text and images to generate meaningful responses. This ability sets it apart from previous AI models like Llama and brings open source closer to mainstream vision models from OpenAI and Google.
The key to Chameleon’s success lies in its fully token-based architecture, allowing it to reason over images and text jointly. Meta’s team overcame technical challenges related to optimization stability and scaling using new methods and training techniques. As a result, Chameleon can handle prompts that require outputs with both text and images seamlessly.
Human evaluations have shown that Chameleon matches or exceeds the performance of models like Gemini Pro and GPT-4V when dealing with mixed sequences of images and text. However, evaluations on interpreting infographics and charts were excluded. The publicly released model can only generate text outputs, with increased safety levels.
Chameleon represents a source of inspiration for researchers exploring alternative ways to train and design AI models. For users, it means being one step closer to having AI assistants that can better understand the context they operate in without relying on closed platforms.
Source link
Source link: https://medium.com/@infoimtiaz2/meta-just-dropped-an-open-source-gpt-4o-style-model-heres-what-it-means-67cffd688873?source=rss——chatgpt-5
GIPHY App Key not set. Please check settings