ERNIE 5.5 Flash Preview

ERNIE 5.5 Flash Preview is Baidu's first post-trained model built on ERNIE 5.5 Flash Base. It is a mixture-of-experts model with 280 billion total parameters and 14 billion active parameters, a 1,048,576-token context window, and text-only input and output. Baidu describes it as strong at coding, 3D modeling, knowledge work, and content creation, and as an intermediate post-training checkpoint: some capabilities may still fall short, and an official (non-preview) release is planned. The API is available on the Qianfan platform as ernie-5.5-flash-preview. Baidu says the weights will be open-sourced under Apache 2.0. Qianfan international pay-as-you-go pricing is $0.28 per million input tokens, $0.028 per million cache-hit tokens, and $1.13 per million output tokens. On Qianfan, thinking cannot be disabled, and reasoning effort cannot be adjusted (the documented default is high).