How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suites
Paper
•
2404.16821
•
Published
•
59
InternVL1.5
A Pioneering Open-Source Alternative to GPT-4V
-
OpenGVLab/InternVL-Chat-V1-5
Image-Text-to-Text • Updated • 2.18k • 416Note Released at 2024.04.18 | 26B parameters | Support 4K image; super strong OCR; Approaching the performance of GPT-4V and Gemini Pro on various benchmarks like MMMU, DocVQA, ChartQA, MathVista, etc. (🔥new)
OpenGVLab/InternViT-6B-448px-V1-5
Image Feature Extraction • 6B • Updated • 644 • 77Note Released at 2024.04.20 | Vision Foundation Model | support dynamic resolution, super strong OCR (🔥new)
OpenGVLab/InternViT-300M-448px
Image Feature Extraction • 0.3B • Updated • 2.76k • 62Note Released at 2024.05.25 | 304M parameters | Distilled small vision foundation model with 300M parameters. (🔥new)
OpenGVLab/InternVL-Chat-V1-5-AWQ
Image-Text-to-Text • 9B • Updated • 477 • 11Note Relased at 2024.05.28 | 26B parameters | The 4bit version of InternVL-Chat-V1-5.
OpenGVLab/Mini-InternVL-Chat-4B-V1-5
Image-Text-to-Text • 4B • Updated • 234 • 62Note Released at 2024.05.28 | 4B parameters | 🚀🚀 Only 4B parameters, anyone can deploy it locally. (🔥new)
OpenGVLab/Mini-InternVL-Chat-2B-V1-5
Image-Text-to-Text • 2B • Updated • 985 • 74Note Released at 2024.05.19 | 2B parameters | 🚀🚀 Only 2B parameters, anyone can deploy it locally. (🔥new)