An open foundation model
Baidu has released ERNIE 5.5 Flash Base as an open-source foundation model. The model has 280 billion parameters and uses a Mixture-of-Experts (MoE) architecture. MoE activates only part of the model for each query, which lowers compute cost.
Pre-trained on 33.5 trillion tokens
The model was pre-trained on 33.5 trillion tokens, meaning it learned from a very large body of text. Baidu ships it deliberately without fine-tuning.
Freedom for developers
An un-fine-tuned base model lets developers shape it with their own data and for their own use cases. Baidu is offering a foundation to build on rather than a finished product.
Why it matters
A major Chinese company sharing a large-scale model with open weights raises competition in the open-source model ecosystem. For developers, it adds a new option they can run on their own infrastructure.
