Ali Tongyi's Qwen3-VL Introduces Two Dense Models, Offering Open-Source Solutions for Diverse Scenarios
2025-10-22 / Read about 0 minute
Author:小编   

On October 22, 2025, Ali Tongyi Qianwen unveiled the expansion of its Qwen3-VL model family with the addition of two dense model variants, namely the 2B and 32B models. These additions are designed to cater to a wide spectrum of visual language understanding scenarios, ranging from lightweight applications to more optimal, well-balanced use cases. The Qwen3-VL family now comes in two distinct versions: the Instruct version, which is characterized by its rapid response times, stable execution, and suitability for dialogue systems and tool invocation; and the Thinking version, which boasts enhanced capabilities in long-chain reasoning and complex visual comprehension, enabling it to engage in "thinking with images."