Baidu releases 280B parameter ERNIE 5.5 Flash Base

Baidu released ERNIE-5.5-Flash-Base, a sparse Mixture of Experts language model, as detailed on its Hugging Face model page. The model has 280 billion total parameters and activates 14 billion parameters per token.

The text only base model was trained from scratch on 33.5 trillion tokens with a training context of up to 128,000 tokens. Fine tuning and inference support context windows of up to 1,048,576 tokens. Baidu released FP8 and BF16 weights for domain adaptation, instruction tuning and reinforcement learning research.

The architecture uses 43 layers, including 42 Mixture of Experts layers with 512 routed experts and one shared expert per layer. Baidu says that under a unified evaluation pipeline the model delivers strong results across knowledge, reasoning, math, code, multilingual and long-context benchmarks. It is released under Baidu’s own ERNIE 5.5 Flash Base Model License. The model supports serving through SGLang and vLLM.