Seed1.5-VL

ByteDance-SeedApache-2.0长期未更新

字节跳动开源的视觉语言基础模型，主打通用多模态理解与推理，在 60 个公开基准中有 38 项达到 SOTA，配套 Notebook 教程方便上手。

Seed1.5-VL, a vision-language foundation model designed to advance general-purpose multimodal understanding and reasoning, achieving state-of-the-art performance on 38 out of 60 public benchmarks.

暂无版本1,592star65fork同步于 2026-11-07 07:31 UTC

视觉语言模型多模态理解多模态推理基础模型基准评测

对比 GitHub 官网 cookbook large-language-model multimodal-large-language-models vision-language-model

怎么上手

由 README 提炼，执行前请确认

原文

获取 API_KEY 后参考各目录示例运行，如 ./GradioDemo

硬件要求按 README 说明

README 未说明本地部署的显存或 GPU 要求；模型通过火山引擎 API 调用，无需本地硬件

下载：源码 zip·GitHub Releases（安装包）本站不镜像安装包

项目说明

由 README 自动翻译提炼

没人工点评？写下你的经验

字节 Seed1.5-VL 多模态基础模型的 API 调用示例集，帮你快速上手视觉理解与推理

字节跳动开源的视觉语言基础模型，主打通用多模态理解与推理，在 60 个公开基准中有 38 项达到 SOTA，配套 Notebook 教程方便上手。

核心能力

提供 Gradio 在线/离线 Demo 示例代码
演示 LongCoT 长链思考的开关用法
覆盖 2D 定位、3D 理解、视频理解示例
包含 GUI Agent 交互任务调用范例
模型在 60 个公开基准中 38 项达 SOTA

容易踩的坑

需先在火山引擎申请 API_KEY 才能调用；仓库只提供调用示例，不含模型权重；模型 ID 为 doubao-1-5-thinking-vision-pro-250428

内容由模型根据 README 生成，可能不完整或不准确，以官方文档为准。看原文 README

用户点评

登录用户的真实使用记录，可按版本和硬件筛选

共 0 条

还没有人点评。第一个说出真实体验的人最有价值。

README 原文

未翻译

在 GitHub 打开

Today, we are excited to introduce Seed1.5-VL 🚀, a powerful and efficient vision-language foundation model designed for advanced general-purpose multimodal understanding and reasoning.

🌟 Highlights

🧠 Efficient Powerhouse: Achieves top performance with a relatively modest architecture, 532M vision encoder & 20B active parameter MoE LLM.
🏆 Exceptional Benchmark Performance: Delivers State-of-the-Art results on 38 out of 60 public VLM benchmarks, demonstrating broad competence.
💡 Versatile Capabilities: Excels across diverse capabilities including complex reasoning (e.g., visual puzzles like Rebus), OCR, diagram understanding, visual grounding, 3D spatial understanding, and video comprehension.
🤖 Advanced Agent-Centric Abilities: Demonstrates leading performance in interactive agent tasks, showcasing strong capabilities in GUI control and gameplay.

This repository offers usage cookbook and best practices designed to help developers effectively use Seed1.5-VL.

📢 News

2025-05-13: We have deployed our Seed1.5-VL on 🤗 HuggingFace Spaces, Welcome to try out our model!
2025-05-12: We have released the Seed1.5-VL Technical Report.
2025-05-12: We are extremely delighted to release the flagship Seed1.5-VL on Volcano Engine. The Model ID is doubao-1-5-thinking-vision-pro-250428. You can try it now!

📮 Notice

Call for Bad Cases: If you have encountered any cases where the model performs poorly, we would greatly appreciate it if you could share them in the issue [https://github.com/ByteDance-Seed/Seed1.5-VL/issues/12]

📖 Seed1.5-VL Cookbook

The Seed1.5-VL cookbook is designed to help you start using the Seed1.5-VL API with diverse code samples. Our flagship Seed1.5-VL has been deployed on Volcano Engine. After obtaining your API_KEY, you can use the examples in this cookbook to rapidly understand and leverage the diverse capabilities of our Seed1.5-VL.

Quick Start

Cookbook for online/offline Gradio Demo
Cookbook for turning on/off LongCoT
Cookbook for 2D Grounding
Cookbook for 3D Understanding
Cookbook for Video Understanding
Cookbook for GUI Agents

Citations

If you Seed1.5-VL useful in your research or applications, please consider giving us a star 🌟 and citing it by the following BibTeX entry.

@article{seed2025seed1_5vl,
  title={Seed1.5-VL Technical Report},
  author={ByteDance Seed Team},
  journal={arXiv preprint arXiv:2505.07062},
  year={2025}
}

License

This repo is under Apache-2.0 License.

About ByteDance Seed Team

Founded in 2023, ByteDance Seed Team is dedicated to crafting the industry's most advanced AI foundation models. The team aspires to become a world-class research team and make significant contributions to the advancement of science and society.