ERNIE 5.5 Flash Base

Baidu's ERNIE-5.5-Flash-Base, the raw end-of-pre-training checkpoint of its new ERNIE 5.5 Flash sparse Mixture-of-Experts language model: 280B total / 14B active parameters, pre-trained from scratch on 33.5T tokens with a 128K training context and 1M-token context supported for fine-tuning and inference. Released with no mid- or post-training so developers can use it as a starting point for domain adaptation, instruction tuning and RL research; FP8 and BF16 weights are on Hugging Face under the ERNIE 5.5 Flash Base Model License Agreement v1.0. Text-only; served via the completions endpoint with SGLang and vLLM. Benchmarks are vendor-reported; Baidu says a post-trained version is to follow.