Phi-4-mini

Available · Language

Phi-4-mini is a 3.8-billion-parameter text model in Microsoft's Phi family of small language models, released on 26 February 2025 with Phi-4-multimodal. It was offered in Azure AI Foundry, the NVIDIA API Catalog and as MIT-licensed open weights on Hugging Face, and supports function calling, 23 languages and a 128K-token context. [1] [2] Primary source

Timeline of Phi-4-mini →

Claims and evidence

  • Released Primary source[1]page date and first paragraph [2]Training, Model: Release date February 2025
  • Status AvailableWeights remain downloadable from Microsoft's Hugging Face organisation; Microsoft Foundry lists the model as GA without a retirement date (checked 2026-10-01). Primary source[2]weights and licence (MIT) [3]Foundry Models from partners and community, Microsoft: Phi-4-mini-instruct, GA, no retirement date
  • Successor of Phi-3.5The technical report calls Phi-3.5-Mini, the mini variant of the Phi-3.5 release, the predecessor of Phi-4-Mini. Primary source[4]Abstract [2]Training, Model: Architecture [5]FAQ: What are the latest Phi models?
  • Change · Languages Vocabulary grows to 200K tokens, which Microsoft says is meant to serve multilingual applications better than Phi-3.5-mini.Compared with Phi-3.5 Primary source[4] [2]Training, Model: Architecture
  • Change · Architecture Adds grouped-query attention and shared input and output embeddings compared with Phi-3.5-mini.Compared with Phi-3.5 Primary source[2]Training, Model: Architecture
  • Change · Tool use Microsoft lists built-in function calling as new compared with Phi-3.5-mini.Compared with Phi-3.5 Primary source[5]FAQ: What are the latest Phi models?
  • Input text Primary source[2]Training, Model: Inputs
  • Output text Primary source[2]Training, Model: Outputs
  • Feature Function calling Primary source[1]What is Phi-4-mini? [2]Usage: Tool-enabled function-calling format
  • Feature MultilingualThe model card lists 23 supported languages. Primary source[2]Training, Model: Supported languages
  • Open weights Yes Primary source[2]licence: MIT
  • Context window 128K tokensThe announcement writes 128,000 tokens. Primary source[2]Training, Model: Context length [1]What is Phi-4-mini?
  • Parameters 3.8B Primary source[1]What is Phi-4-mini? [2]Training, Model: Architecture
  • Access API, Open-weights download, cloud partnerAt launch: Azure AI Foundry (api), Hugging Face (open-weights-download) and the NVIDIA API Catalog (cloud-partner). Primary source[1]first paragraph; Learn more about Phi-4

Lineage

Predecessors

Successors

No known successor.

Based on

Not derived from another model.

Variants and derived

Siblings

None recorded.

All ancestors

All descendants

Variants

No variants recorded in this record.

Related AI Radar coverage

AI Radar coverage starts in June 2026; no coverage linked yet.

All model releases from Microsoft on AI Radar →