Back to Models
Microsoft
Phi 3
Open source3 variants
Microsoft's 3.8B Phi-3 Mini instruct model with 128K context — remarkable reasoning for its size, ideal for edge and mobile.
131K tokensFree / Open weightsTransformerMIT
GAIA
Phi-3-mini-128k-instruct
13.9%GAIA-LLM-Agent w/ Phi-3 (proxy score)
instruct3.8B
Microsoft's 3.8B Phi-3 Mini instruct model with 4K context — same strong reasoning as the 128k variant in a smaller memory footprint.
4K tokensFree / Open weights
instruct4.2B
Microsoft's 4.2B multimodal Phi-3 model — combines vision and language with 128K context for efficient on-device image understanding.
131K tokensFree / Open weights
