Skip to content

Phi-4 Mini 3.8B

by Microsoft · phi family

3.8B

parameters

text-generation code-generation reasoning math summarization

Phi-4 Mini 3.8B is Microsoft's compact reasoning model with 128K context, punching far above its weight on math and reasoning benchmarks. It continues the Phi tradition of proving that smaller, carefully trained models can compete with much larger ones. At Q4 it fits on any 8 GB GPU or Mac with minimal overhead — an excellent choice for users who want strong reasoning capabilities on resource-constrained hardware. Great for math, coding, and structured tasks.

Quick Start with Ollama

ollama run 3.8b-q4_K_M
Resources Ollama Hugging Face Official Page
Creator Microsoft
Parameters 3.8B
Architecture transformer-decoder
Context 128K tokens
Released Feb 26, 2025
License MIT
Ollama phi4-mini

Quantization Options

Format File Size VRAM Required Quality Ollama Tag
Q4_K_M rec 2.5 GB 4.5 GB 3.8b-q4_K_M
Q8_0 4.2 GB 6.5 GB 3.8b-q8_0
F16 7.8 GB 10.5 GB 3.8b-fp16

Compatible Hardware

Q4_K_M requires 4.5 GB VRAM

Compatible Hardware

HardwareVRAMTypeFitEst. Speed
Mac Studio M4 Ultra 512GB512 GBmacRuns~182 tok/s
Mac Pro M2 Ultra 192GB192 GBmacRuns~178 tok/s
Mac Studio M4 Ultra 192GB192 GBmacRuns~182 tok/s
Mac Studio M4 Max 128GB128 GBmacRuns~121 tok/s
MacBook Pro M4 Max 128GB128 GBmacRuns~121 tok/s
MacBook Pro M5 Max 128GB128 GBmacRuns~121 tok/s
NVIDIA RTX PRO 6000 Blackwell96 GBgpuRuns~427 tok/s
MacBook Pro M3 Max 96GB96 GBmacRuns~89 tok/s
Mac mini M4 Pro 64GB64 GBmacRuns~61 tok/s
Mac Studio M4 Max 64GB64 GBmacRuns~121 tok/s
MacBook Pro M4 Max 64GB64 GBmacRuns~121 tok/s
MacBook Pro M5 Max 64GB64 GBmacRuns~121 tok/s
NVIDIA RTX 6000 Ada Generation48 GBgpuRuns~213 tok/s
NVIDIA RTX A600048 GBgpuRuns~171 tok/s
NVIDIA RTX PRO 5000 Blackwell48 GBgpuRuns~213 tok/s
Mac mini M4 Pro 48GB48 GBmacRuns~61 tok/s
MacBook Pro M3 Max 48GB48 GBmacRuns~89 tok/s
MacBook Pro M4 Max 48GB48 GBmacRuns~121 tok/s
MacBook Pro M4 Pro 48GB48 GBmacRuns~61 tok/s
MacBook Pro M5 Max 48GB48 GBmacRuns~91 tok/s
MacBook Pro M5 Pro 48GB48 GBmacRuns~61 tok/s
Mac Studio M4 Max 36GB36 GBmacRuns~121 tok/s
MacBook Pro M3 Pro 36GB36 GBmacRuns~33 tok/s
MacBook Pro M5 Max 36GB36 GBmacRuns~91 tok/s
NVIDIA RTX 5000 Ada Generation32 GBgpuRuns~160 tok/s
NVIDIA GeForce RTX 509032 GBgpuRuns~398 tok/s
iMac M4 32GB32 GBmacRuns~27 tok/s
Mac mini M4 32GB32 GBmacRuns~27 tok/s
MacBook Air M4 32GB32 GBmacRuns~27 tok/s
MacBook Air M5 32GB32 GBmacRuns~27 tok/s
MacBook Pro M5 32GB32 GBmacRuns~27 tok/s
AMD Radeon RX 7900 XTX24 GBgpuRuns~213 tok/s
NVIDIA GeForce RTX 309024 GBgpuRuns~208 tok/s
NVIDIA GeForce RTX 3090 Ti24 GBgpuRuns~224 tok/s
NVIDIA GeForce RTX 409024 GBgpuRuns~224 tok/s
NVIDIA RTX A500024 GBgpuRuns~171 tok/s
iMac M3 24GB24 GBmacRuns~22 tok/s
Mac mini M2 24GB24 GBmacRuns~22 tok/s
Mac mini M4 Pro 24GB24 GBmacRuns~61 tok/s
MacBook Air M2 24GB24 GBmacRuns~22 tok/s
MacBook Air M4 24GB24 GBmacRuns~27 tok/s
MacBook Air M5 24GB24 GBmacRuns~27 tok/s
MacBook Pro M4 Pro 24GB24 GBmacRuns~61 tok/s
MacBook Pro M5 24GB24 GBmacRuns~27 tok/s
MacBook Pro M5 Pro 24GB24 GBmacRuns~61 tok/s
AMD Radeon RX 7900 XT20 GBgpuRuns~178 tok/s
NVIDIA RTX 4000 Ada Generation20 GBgpuRuns~80 tok/s
MacBook Pro M3 Pro 18GB18 GBmacRuns~33 tok/s
AMD Radeon RX 6900 XT16 GBgpuRuns~114 tok/s
AMD Radeon RX 6800 XT16 GBgpuRuns~114 tok/s
AMD Radeon RX 7800 XT16 GBgpuRuns~139 tok/s
AMD Radeon RX 9060 XT 16GB16 GBgpuRuns~120 tok/s
AMD Radeon RX 9070 XT16 GBgpuRuns~144 tok/s
AMD Radeon RX 907016 GBgpuRuns~120 tok/s
Intel Arc A77016 GBgpuRuns~124 tok/s
NVIDIA GeForce RTX 4060 Ti 16GB16 GBgpuRuns~64 tok/s
NVIDIA GeForce RTX 4070 Ti Super16 GBgpuRuns~149 tok/s
NVIDIA GeForce RTX 4080 Super16 GBgpuRuns~164 tok/s
NVIDIA GeForce RTX 408016 GBgpuRuns~159 tok/s
NVIDIA GeForce RTX 5060 Ti 16GB16 GBgpuRuns~100 tok/s
NVIDIA GeForce RTX 5070 Ti16 GBgpuRuns~199 tok/s
NVIDIA GeForce RTX 508016 GBgpuRuns~213 tok/s
NVIDIA RTX A400016 GBgpuRuns~100 tok/s
iMac M1 16GB16 GBmacRuns~15 tok/s
iMac M4 16GB16 GBmacRuns~27 tok/s
Mac mini M1 16GB16 GBmacRuns~15 tok/s
Mac mini M4 16GB16 GBmacRuns~27 tok/s
MacBook Air M2 16GB16 GBmacRuns~22 tok/s
MacBook Air M3 16GB16 GBmacRuns~22 tok/s
MacBook Air M4 16GB16 GBmacRuns~27 tok/s
MacBook Air M5 16GB16 GBmacRuns~27 tok/s
MacBook Pro M1 16GB16 GBmacRuns~15 tok/s
MacBook Pro M2 Pro 16GB16 GBmacRuns~44 tok/s
MacBook Pro M5 16GB16 GBmacRuns~27 tok/s
AMD Radeon RX 6700 XT12 GBgpuRuns~85 tok/s
AMD Radeon RX 7700 XT12 GBgpuRuns~96 tok/s
Intel Arc B58012 GBgpuRuns~101 tok/s
NVIDIA GeForce RTX 3060 12GB12 GBgpuRuns~80 tok/s
NVIDIA GeForce RTX 3080 12GB12 GBgpuRuns~203 tok/s
NVIDIA GeForce RTX 4070 Super12 GBgpuRuns~112 tok/s
NVIDIA GeForce RTX 4070 Ti12 GBgpuRuns~112 tok/s
NVIDIA GeForce RTX 407012 GBgpuRuns~112 tok/s
NVIDIA GeForce RTX 507012 GBgpuRuns~149 tok/s
NVIDIA GeForce GTX 1080 Ti11 GBgpuRuns~108 tok/s
NVIDIA GeForce RTX 2080 Ti11 GBgpuRuns~137 tok/s
Intel Arc B57010 GBgpuRuns~84 tok/s
NVIDIA GeForce RTX 3080 10GB10 GBgpuRuns~169 tok/s
AMD Radeon RX 6600 XT8 GBgpuRuns~57 tok/s
AMD Radeon RX 76008 GBgpuRuns~64 tok/s
AMD Radeon RX 9060 XT 8GB8 GBgpuRuns~60 tok/s
Intel Arc A7508 GBgpuRuns~114 tok/s
NVIDIA GeForce GTX 10708 GBgpuRuns~57 tok/s
NVIDIA GeForce RTX 2060 Super8 GBgpuRuns~100 tok/s
NVIDIA GeForce RTX 2070 Super8 GBgpuRuns~100 tok/s
NVIDIA GeForce RTX 2080 Super8 GBgpuRuns~110 tok/s
NVIDIA GeForce RTX 30508 GBgpuRuns~50 tok/s
NVIDIA GeForce RTX 30708 GBgpuRuns~100 tok/s
NVIDIA GeForce RTX 3060 Ti8 GBgpuRuns~100 tok/s
NVIDIA GeForce RTX 4060 Ti 8GB8 GBgpuRuns~64 tok/s
NVIDIA GeForce RTX 40608 GBgpuRuns~60 tok/s
NVIDIA GeForce RTX 50508 GBgpuRuns~50 tok/s
NVIDIA GeForce RTX 5060 Ti 8GB8 GBgpuRuns~100 tok/s
NVIDIA GeForce RTX 50608 GBgpuRuns~75 tok/s
MacBook Air M1 8GB8 GBmacRuns~15 tok/s
MacBook Air M2 8GB8 GBmacRuns~22 tok/s
NVIDIA GeForce GTX 1660 Super6 GBgpuRuns~75 tok/s
NVIDIA GeForce RTX 20606 GBgpuRuns~75 tok/s

Benchmark Scores

70.0
mmlu