Baidu releases ERNIE 5.5 Flash Base weights
A 280B mixture-of-experts base model, 14B active, with weights from the end of pre-training and a 1M-token context for fine-tuning.
Official source (opens in a new tab)
Baidu's 5 November 2026 release page publishes ERNIE 5.5 Flash Base, a sparse mixture-of-experts language model with 280 billion total parameters and 14 billion active per token. It was pre-trained on 33.5 trillion tokens, with a training context of up to 128K tokens, and it supports a 1 million token context for fine-tuning and inference. Baidu says the weights are the checkpoint at the end of pre-training, before mid-training or post-training, and are meant as a starting point for domain adaptation, instruction tuning, and reinforcement-learning research. The page links the weights on Hugging Face, including a BF16 set.
Who it's for: a lab or a developer who wants a base checkpoint to train further, not a finished chat model.
Fits: Best for Writing stays Claude. Best for Vibe Coding stays Cursor. A pre-training checkpoint is not a product on either list. The coding guide is unchanged. The verdict is unchanged.