Model release · Baidu
ERNIE 5.5 Flash Base
- Released
- Review updated
This is a pre-trained base model, not a chat model: the weights come straight from pre-training, with no mid- or post-training. Teams can compare their own fine-tuning methods from one checkpoint and shape it around their tasks. It was trained on 33.5T tokens with a training context of up to 128K, and supports 1M tokens for fine-tuning and inference.
Availability
The weights are on Hugging Face in FP8 and BF16 (baidu/ERNIE-5.5-Flash-Base and the -BF16 repo). The model is text-only and supports up to 1M tokens of context for fine-tuning and inference. It uses Baidu's own ERNIE 5.5 Flash Base License v1.0. Non-commercial use is free. Commercial use is free of license fees up to US$50M in annual related revenue if you display "Powered by ERNIE" and send a one-time notice; above that you need Baidu's written authorization first. The license does not cover the EU, the UK or South Korea. As a base model it is served through the completions endpoint, not chat. Baidu does not mention an API or Qianfan, and says a post-trained ERNIE 5.5 Flash will follow.