Microsoft Unveils Cutting-Edge Phi 3.5 AI Models

New models offer advanced capabilities in natural language processing, multimodal AI, and high-performance computing, available under an open-source MIT license.

Phi 3.5 Mini Instruct excels in handling long documents and complex reasoning with 128K context length.
Phi 3.5 MoE leverages a Mixture of Experts architecture for scalable and efficient performance.
Phi 3.5 Vision Instruct integrates text and image processing for advanced multimodal tasks.
All models demonstrate near state-of-the-art performance in various benchmarks, outperforming competitors.
Available on Hugging Face, these models are open for commercial use and modification.