DeepSeek, the viral AI company, has released a new set of multimodal AI models that it claims can outperform OpenAI’s DALL-E 3.
The models, which are available for download from the AI dev platform Hugging Face, are part of a new model family that DeepSeek is calling Janus-Pro. They range in size from 1 billion to 7 billion parameters. Parameters roughly correspond to a model’s problem-solving skills, and models with more parameters generally perform better than those with fewer parameters.
Janus-Pro is under an MIT license, meaning it can be used commercially without restriction.
As far as I understand, the training data is closed source. But, the methodology of training is open source which allows independent parties to recreate the model from scratch and see similar results. Not only can you download the full >400GB model using huggingface or ollama, but they also offer distilled versions of the model which are small enough to run on something like a raspberry pi. i’m running it locally on my machine at home with perplexica (perplexity.ai lookalike with searching capabilities)