Alibaba has released the QVQ-72B-Preview, a new visual reasoning model designed for natural language processing (NLP) and multimodal tasks. The model, which has been made available on platforms such as Kaggle and SiliconFlow, is noted for its performance in benchmarks like MMLU and VQAv2, reportedly surpassing leading models including GPT-4. The QVQ-72B-Preview is licensed under Apache 2 and is part of Alibaba's ongoing efforts to enhance AI capabilities. Additionally, a Turbo version of the model has been launched, further emphasizing its advanced features. Users are encouraged to explore the model on various platforms, with additional benchmarks and comparisons to previous models expected to follow.