microsoft/Phi-3-vision-128k-instruct

About Model

As part of the Phi-3 model family, Phi-3-vision-128k-instruct combines text and vision modalities to perform complex reasoning tasks across large contexts. The model is trained on high-quality, reasoning-rich datasets, including synthetic data and filtered publicly available web content. It has undergone supervised fine-tuning and direct preference optimization to enhance instruction-following accuracy and safety. Phi-3-vision-128k-instruct excels in tasks such as image captioning, visual question answering, and document analysis. The model is available on platforms like Hugging Face and Azure AI Foundry, supporting applications that require deep multimodal comprehension.