DataLearner logoDataLearnerAI
Latest AI Insights
Model Evaluations
Model Directory
Model Comparison
Resource Center
Tools

加载中...

DataLearner logoDataLearner AI

A knowledge platform focused on LLM benchmarking, datasets, and practical instruction with continuously updated capability maps.

Products

  • Leaderboards
  • Model comparison
  • Datasets

Resources

  • Tutorials
  • Editorial
  • Tool directory

Company

  • About
  • Privacy policy
  • Data methodology
  • Contact

© 2026 DataLearner AI. DataLearner curates industry data and case studies so researchers, enterprises, and developers can rely on trustworthy intelligence.

Privacy policyTerms of service
Page navigation
目录
Model catalogQwen3-VL-8B-Instruct
QW

Qwen3-VL-8B-Instruct

Qwen3-VL-8B-Instruct

Release date: 2025-10-15更新于: 2025-10-15 08:23:371,009
Live demoGitHubHugging FaceCompare
Parameters
88.0亿
Context length
256K
Chinese support
Not supported
Reasoning ability

Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology

Qwen3-VL-8B-Instruct

Model basics

Reasoning traces
Not supported
Thinking modes
Thinking modes not supported
Context length
256K tokens
Max output length
No data
Model type
多模态大模型
Release date
2025-10-15
Model file size
No data
MoE architecture
No
Total params / Active params
88.0B / N/A
Knowledge cutoff
No data
Qwen3-VL-8B-Instruct

Open source & experience

Code license
Apache 2.0
Weights license
Apache 2.0- 免费商用授权
GitHub repo
https://github.com/QwenLM/Qwen3-VL
Hugging Face
https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct
Live demo
No live demo
Qwen3-VL-8B-Instruct

Official resources

Paper
Qwen3 Technical Report
DataLearnerAI blog
No blog post yet
Qwen3-VL-8B-Instruct

API details

API speed
3/5
No public API pricing yet.
Qwen3-VL-8B-Instruct

Benchmark Results

Qwen3-VL-8B-Instruct currently shows benchmark results led by DocVQA (2 / 5, score 96.10), MMMU (23 / 27, score 69.60). This page also consolidates core specs, context limits, and API pricing so you can evaluate the model from benchmark results and deployment constraints together.

Thinking
All modesNormal

多模态理解

2 evaluations
Benchmark / mode
Score
Rank/total
DocVQA
Off
96.10
2 / 5
MMMU
Off
69.60
23 / 27
View benchmark analysisCompare with other models
Qwen3-VL-8B-Instruct

Publisher

阿里巴巴
阿里巴巴
View publisher details
Qwen3-VL-8B-Instruct

Model Overview

Qwen3-VL 8B 简介

Qwen3-VL 8B 属于 Qwen3-VL 系列的中型开源权重，面向图像/视频/文本的统一理解与推理。根据官方模型卡与说明，Qwen3-VL 在本代提供了更长的上下文支持（原生 256K，扩展至 1M）、更强的视觉-文本对齐（DeepStack 多层特征融合）与时间建模（文本-时间戳对齐），并在 GUI 要素识别与工具调用等“视觉代理”场景中提供能力示例。

架构与技术要点

  • 关键组件：Interleaved-MRoPE 位置编码（覆盖时间/宽度/高度频带）、DeepStack 视觉特征融合、文本-时间戳对齐以增强视频时序定位。
  • 上下文窗口：原生 256K，可扩展至 1M。
  • 模态与能力：支持图像/视频作为输入，输出为文本；覆盖 OCR（32 种语言）、版面结构解析、空间关系/遮挡判断、时序理解与长文档/长视频检索定位等。
  • 许可与获取：Apache-2.0 许可；权重与使用指南在 Hugging Face 提供。

参数与资源

模型卡显示参数量约 8.77B；提供 Transformers 直接调用与推理超参示例。

DataLearner 官方微信

欢迎关注 DataLearner 官方微信，获得最新 AI 技术推送

DataLearner 官方微信二维码