Major models like Google's Gemma, Meta's Llama, and OpenAI's GPT-2 are released with open weights and sometimes inference-time code. The Open Source Institute defines true open-source AI as requiring training code and detailed data information for replicability and transparency. This allows researchers to better understand biases and limitations in models. Companies like Elon Musk's xAI and HuggingFace have navigated open sourcing, with xAI releasing some systems while keeping others proprietary, indicating a nuanced approach in the evolving AI landscape.
A fully open source release, including training code, can give researchers more visibility into how a model works at a core level, potentially revealing biases or limitations that are inherent to the model's architecture instead of its parameter weights.
It's currently unclear whether DeepSeek's planned open source release will also include the code the team used when training the model. That kind of training code is necessary to meet the Open Source Institute's formal definition of 'Open Source AI'.
Collection
[
|
...
]