Phi-3.5-vision-instruct

Phi-3.5-vision-instruct is an open multimodal model with 4.2 billion parameters and support for up to 128K token context window. The model specializes in understanding and analyzing multiple image frames, improving performance on single-image benchmarks while enabling multi-image comparison, summarization, and even video analysis. The model underwent post-training for safety to improve instruction following, alignment, and robust handling of visual and text inputs, and is released under the MIT license.