Qwen3-VL-4B-Thinking
Qwen3-VL-4B-Thinking is an AI model published by 阿里巴巴, released on 2025-10-15, for 推理大模型, with 44.4B parameters, and 256K tokens context length, under the Apache 2.0 license.
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
Qwen3-VL-4B-Thinking 是阿里巴巴 Qwen 团队在 Qwen3-VL 系列下发布的推理增强多模态模型,支持图像/视频作为输入并输出文本。4B/8B(Instruct 与 Thinking)于 2025-10-15 公告上线。
覆盖 OCR(32 种语言)、版面/结构化文档理解、空间/遮挡关系与 2D/3D grounding、长视频事件索引与定位、GUI 元素识别与步骤性交互等。
欢迎关注 DataLearner 官方微信,获得最新 AI 技术推送
