Microsoft debuts Phi-4, a new generative AI model, in research preview

Microsoft unveils Phi-4, a 14 billion parameter AI model, available on Azure AI Foundry.

: Microsoft has introduced Phi-4, a new generative AI model in its Phi series, available through limited access on Azure AI Foundry. This model, boasting 14 billion parameters, advances in areas such as math problem solving due to enhanced training data quality. It rivals other small models like GPT-4o mini and Claude 3.5 Haiku. Phi-4 marks the first release after Sébastien Bubeck's departure from Microsoft.

Microsoft has recently introduced Phi-4, the latest in their line of generative AI models, emphasizing improvements in math problem solving. The model is currently accessible in a limited capacity exclusively through Microsoft's Azure AI Foundry for research under a specific license agreement.

Phi-4 is a 14 billion parameter small language model, designed to compete with others like GPT-4o mini and Claude 3.5 Haiku. The model's enhanced performance is attributed to high-quality synthetic datasets and post-training improvements, reflecting the growing focus on such innovations in the AI field.

This release follows major developments in the AI community, including the departure of key figure Sébastien Bubeck to OpenAI. His influence had been pivotal to previous Phi models, making Phi-4 significant as the first model launched post his tenure at Microsoft.

Generative AIPhi4Data Quality