Microsoft has introduced Phi-3 Mini, a lightweight AI model that can understand complex instructions like larger models.

The model is reportedly able to process 3.8 billion parameters and is available in two context-length 4K and 128K tokens.

Additionally, Microsoft plans to release two versions of this small model, Phi-3 Small and Phi-3 Medium, which are capable of processing 7 billion and 14 billion parameters respectively.

Microsoft claims that the model performs better than Phi-2, a small language model launched in December last year (and larger models like GPT-3.5), is cheaper, and works efficiently on personal devices. Additionally, the model is available on Azure, Hugging Face, and Ollama.