According to CoinMeta, Baidu has open-sourced its Wenxin 5.5 model, which contains 280 billion parameters, for the first time in 16 months. This model is primarily designed for optimizing inference efficiency. It has been pre-trained using 33.5 trillion parameters and supports an inference context of up to 1 million parameters. Baidu has also redesigned its long-text processing mechanism to reduce memory and computational costs. The model made available this time is a pure-text model that has not undergone post-training modifications, intended for developers to perform domain-specific fine-tuning and reinforcement learning. The license for Wenxin 5.5 is more stringent than that of its predecessor; it does not authorize use in the European Union, the United Kingdom, or South Korea. Related products or services have generated revenue of over 50 million US dollars in the past 12 months, and commercial use requires written authorization from Baidu.