Phi-3-medium

Microsoft · · 133 days after Phi-2

Verified

3 of 3 published values traced to a source2 primary sources, 2 archived1 not researched yet

Why this status

Split from the former "phi-3" record, which was named for the family while holding one variant's figures. Evidence: "Phi-3-small (7B) and Phi-3-medium (14B) will be available in the Azure AI model catalog and other model gardens shortly. The cited model card is Phi-3-medium-128k-instruct, which is where the 128K context window comes from." Capabilities evidenced from the primary sources: coding: “Phi-3 models are the most capable and cost-effective small language models (SLMs) available, outperforming models of the same size and next size up across a variety of la”.

What these states mean · How records are judged · Has this record changed?

The largest Phi-3 model at 14 billion parameters, with a 128K-token context variant.

codingsmall-efficient

Details

ReleasedApril 23, 2024
stated in azure.microsoft.com

Phi-3-medium was released on April 23, 2024.

CompanyMicrosoft
FamilyPhi
Released alongsidePhi-3-mini · Phi-3-small
EraMultimodal & open weights
TypeLanguage model · Small language model
ModalitiesText → Text
CapabilitiesCoding
WeightsProprietary
LicenceNot applicable — proprietary
Context window131K tokens
stated in azure.microsoft.com, huggingface.co

Phi-3-medium supports a context window of 131,072 tokens.

Parameters14B
stated in azure.microsoft.com, huggingface.co

Phi-3-medium has 14,000,000,000 parameters.

Current information, elsewhere

This record is a historical one: every figure above is traced to a dated primary source. For what is true today — live pricing, current rankings, availability — these are better maintained than anything here could be.

Sources

Record status: verified · confidence 90/100 · 2 of 2 primary

Split from the former "phi-3" record, which was named for the family while holding one variant's figures. Evidence: "Phi-3-small (7B) and Phi-3-medium (14B) will be available in the Azure AI model catalog and other model gardens shortly. The cited model card is Phi-3-medium-128k-instruct, which is where the 128K context window comes from." Capabilities evidenced from the primary sources: coding: “Phi-3 models are the most capable and cost-effective small language models (SLMs) available, outperforming models of the same size and next size up across a variety of la”.

Primary means published by the organisation that made the model. Verified records require at least one.

Phi family

  1. Phi-2 December 12, 2023
  2. Phi-3-mini April 23, 2024
  3. Phi-3-small April 23, 2024
  4. Phi-3-medium April 23, 2024

Browse models

Compare Phi-3-medium with another model →
See Phi-3-medium on the interactive timeline →