Baidu Releases ERNIE 5.5 Flash Base Model, 280B Parameters
Baidu has released ERNIE 5.5 Flash Base, a large base model that was published before any mid- or post-training. The model uses a sparse design with 280 billion total parameters, of which 14 billion are active at a time. It was trained on 33.5 trillion tokens.
Quick glossaryBaidu, base model, sparse design
- Baidu
- Baidu is a Chinese internet company best known for its search engine and its ERNIE family of AI models.
- base model
- An AI model trained on huge amounts of text, before it is tuned to follow instructions or chat politely.
- sparse design
- A design where only a small part of a model's total size works on each task, which saves computing power.
Burning Potato score: 55%