See how Baidu's models evolved by answering the same challenge across generations.
Provider of advanced AI models and technologies.
2
Jun 2025 – Oct 2025
Baidu's lightweight reasoning MoE model (21B total, 3B active per token) refined for deep multi-step reasoning across logic, math, science, and code. 128K context, Apache-2.0.
Baidu's flagship 300B-parameter text Mixture-of-Experts model with 47B active parameters per token. 123K context, Apache-2.0 open-source. Part of the ERNIE 4.5 series.