Baidu Intelligent Cloud announces the open source of the Qianfan-VL series models.

Recently, Baidu Intelligent Cloud Qianfan announced the launch of a new visual understanding model - Qianfan-VL, which is fully open source. It is reported that this series of models includes three different sizes: 3B, 8B, and 70B, which are deep-optimized visual understanding large models for enterprise-level multimodal application scenarios. Starting now until October 10th, users can experience the 8B and 70B models for free on the Baidu Intelligent Cloud Qianfan platform. Qianfan-VL not only has excellent basic general capabilities but also has been specifically strengthened for high-frequency demands in industrial applications, such as OCR and education vertical scenes, making it perform more outstandingly in practical applications. It is reported that the Qianfan-VL series models are developed by the Baidu Intelligent Cloud Qianfan model research and development team, based on open source models, and completed the entire computational tasks on Baidu's self-developed Kunlun chip P800. The Kunlun chip P800 provides powerful computational support to ensure that the model can efficiently handle massive amounts of data and complex algorithms, while also supporting parallel computing at a scale of 5000 cards for a single task.