The company said that the model was trained on 15 trillion mixed visual and text tokens.

No comments yet…

Login to comment.