Baidu Open-Sources 280-Billion-Parameter ERNIE 5.5 Flash Base Foundation Model
Baidu has released the open-source foundation model ERNIE-5.5-Flash-Base, deliberately without fine-tuning or post-training, leaving developers to decide the subsequent direction themselves, with post-trained versions to be released later. The model uses a MoE architecture, with 280 billion total parameters and 14 billion activated per input, pre-trained on 33.5 trillion tokens, with a training context of up to 128,000 tokens, and is designed to support fine-tuning and inference-time context of up to 1 million tokens. Baidu says it competes on reasoning, code, and long-context benchmarks with same-scale open-source foundation models such as DeepSeek V4 Flash and MiMo V2.5, and the weights are already available on Hugging Face in FP8 and BF16.
Why read this
AI score 79
Baidu's open-source foundation model provides only pre-trained weights, allowing readers to judge what post-training work developers still need to complete themselves after obtaining it.
Topics
Body
Baidu's AI development is still alive, but most of the work is on you. The Chinese company has released ERNIE-5.5-Flash-Base , an open foundation model that ships deliberately without fine-tuning or post-training. Developers are expected to decide for themselves where to take the model. A post-trained version is supposed to follow.
ERNIE 5.5 uses a mixture-of-experts architecture with 280 billion parameters, 14 billion of which are active per input. The model was pre-trained on 33.5 trillion tokens with a training context of up to 128,000 tokens, though it's designed to support a context of up to one million tokens during fine-tuning and inference.
In benchmarks for reasoning, code, and long-context tasks, the model competes with similarly sized open foundation models like DeepSeek V4 Flash or MiMo V2.5, according to Baidu. The weights are available on Hugging Face in FP8 and BF16.