FitMyLLM
Microsoft/Dense

MicrosoftPhi-4-mini 3.8B

Phi-4 Mini — Microsoft's latest small model with strong reasoning and coding.

chatcodingreasoningThinkingTool Use
3.8B
Parameters
125K
Context length
9
Benchmarks
6
Quantizations
1.5M
HF downloads
Architecture
Dense
Released
2025-02-27
Layers
32
KV Heads
8
Head Dim
128
Family
phi

Quantization Options

Context length:
QuantBitsVRAM @ 16KQuality
Q4_K_M4.89
4.3 GB
2.8 + 1.5 KV
good
Q5_K_S5.57
4.6 GB
3.1 + 1.5 KV
good
Q5_K_M5.7
4.7 GB
3.2 + 1.5 KV
good
Q6_K6.56
5.1 GB
3.6 + 1.5 KV
excellent
Q8_08.5
6.0 GB
4.5 + 1.5 KV
lossless
FP1616
9.6 GB
8.1 + 1.5 KV
lossless

Select your GPU above to see speed estimates and compatibility for each quantization.

Deploying for a team or in production? Size GPUs, cost & scaling in Enterprise →
▸ READY TO RUN THIS?RENT BY THE HOUR

RENT A GPU AND RUN PHI-4-MINI 3.8B NOW

Rent on Vast.ai →Or RunPod →

Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.

Community Ratings

Loading ratings...

Benchmarks (9)

Arena Elo1465
HumanEval80.5
MATH71.2
IFEval66.8
MMLU-PRO52.8
BBH46.5
BigCodeBench33.1
GPQA24.6
MUSR16.8

Run this model

▸Easiest way to get started·Beginners
DOCS ↗
curl -fsSL https://ollama.com/install.sh | sh
$ollama run phi4-mini:3.8b-q4_K_M

Downloads and runs automatically. Add --verbose for speed stats.

▸ SETUP GUIDE
>_

Auto-setup with fitmyllm CLI

Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.

pip install fitmyllmthen run fitmyllmLearn more
Auto-detect GPULive tok/s in chatSpeed benchmarks9 inference engines
HuggingFaceOllama LibraryGGUF DownloadsBuild Hardware

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

NVIDIA Tesla C2050
3 GB VRAM • 144 GB/s
NVIDIA
NVIDIA Tesla M2050
3 GB VRAM • 148 GB/s
NVIDIA
NVIDIA Tesla S2050
3 GB VRAM • 148 GB/s
NVIDIA
NVIDIA GeForce GTX 670MX
3 GB VRAM • 67 GB/s
NVIDIA
AmazonNeweggeBayPCPartPicker
AMD Radeon HD 7950
3 GB VRAM • 240 GB/s
AMD
AmazonNeweggeBayPCPartPicker
AMD Radeon HD 7950 Boost
3 GB VRAM • 240 GB/s
AMD
AmazonNeweggeBayPCPartPicker
AMD Radeon HD 7950 Monica BIOS 1
3 GB VRAM • 240 GB/s
AMD
AmazonNeweggeBayPCPartPicker
AMD Radeon HD 7950 Monica BIOS 2
3 GB VRAM • 240 GB/s
AMD
AmazonNeweggeBayPCPartPicker
AMD Radeon HD 7970
3 GB VRAM • 264 GB/s
AMD
AmazonNeweggeBayPCPartPicker
AMD Radeon HD 7970 GHz Edition
3 GB VRAM • 288 GB/s
AMD
AmazonNeweggeBayPCPartPicker
AMD Radeon HD 7970 X2
3 GB VRAM • 264 GB/s
AMD
AmazonNeweggeBayPCPartPicker
NVIDIA GeForce GTX 770M
3 GB VRAM • 96 GB/s
NVIDIA
AmazonNeweggeBayPCPartPicker
NVIDIA GeForce GTX 780
3 GB VRAM • 288 GB/s
NVIDIA
AmazonNeweggeBayPCPartPicker
NVIDIA GeForce GTX 780 Rev. 2
3 GB VRAM • 288 GB/s
NVIDIA
AmazonNeweggeBayPCPartPicker
NVIDIA GeForce GTX 780 Ti
3 GB VRAM • 337 GB/s
NVIDIA
AmazonNeweggeBayPCPartPicker
AMD Radeon HD 7950 Mac Edition
3 GB VRAM • 240 GB/s
AMD
AmazonNeweggeBayPCPartPicker
AMD Radeon HD 7990
3 GB VRAM • 288 GB/s
AMD
AmazonNeweggeBayPCPartPicker
AMD Radeon HD 8950 OEM
3 GB VRAM • 240 GB/s
AMD
AmazonNeweggeBayPCPartPicker
AMD Radeon HD 8970 OEM
3 GB VRAM • 264 GB/s
AMD
AmazonNeweggeBayPCPartPicker
AMD Radeon HD 8990 OEM
3 GB VRAM • 288 GB/s
AMD
AmazonNeweggeBayPCPartPicker
NVIDIA GeForce GTX 870M
3 GB VRAM • 120 GB/s
NVIDIA
AmazonNeweggeBayPCPartPicker
AMD Radeon R9 280
3 GB VRAM • 240 GB/s
AMD
AmazonNeweggeBayPCPartPicker
NVIDIA GeForce GTX 1060 3 GB
3 GB VRAM • 192 GB/s
NVIDIA
AmazonNeweggeBayPCPartPicker
NVIDIA
AmazonNeweggeBayPCPartPicker
NVIDIA P106-090
3 GB VRAM • 192 GB/s
NVIDIA
AmazonNeweggeBayPCPartPicker
NVIDIA GeForce GTX 1050 3 GB
3 GB VRAM • 84 GB/s
NVIDIA
AmazonNeweggeBayPCPartPicker
AMD Radeon RX 5300M
3 GB VRAM • 168 GB/s
AMD
AmazonNeweggeBayPCPartPicker
AMD Radeon RX 5300 OEM
3 GB VRAM • 168 GB/s
AMD
AmazonNeweggeBayPCPartPicker
NVIDIA GeForce GT 555M 3 GB
3 GB VRAM • 43.2 GB/s
NVIDIA
AmazonNeweggeBayPCPartPicker
AMD FirePro S10000
3 GB VRAM • 240 GB/s
AMD
AmazonNeweggeBayPCPartPicker

Find the best GPU for Phi-4-mini 3.8B

Build Hardware for Phi-4-mini 3.8B

Model Card

View on HuggingFace

Phi-4 Mini — Microsoft's latest small model with strong reasoning and coding.

▸ SPEC SHEET

Phi-4-mini 3.8B — 3.8B Dense.

▸ SPECIFICATIONS
PARAMETERS
3.8B
ARCHITECTURE
Dense Transformer
CONTEXT LENGTH
125K tokens
CAPABILITIES
chat, coding, reasoning
RELEASE DATE
2025-02-27
PROVIDER
Microsoft
FAMILY
phi
▸ VRAM REQUIREMENTS
QUANTBPWVRAMQUALITY
Q4_K_M4.892.8 GB94%
Q5_K_S5.573.1 GB96%
Q5_K_M5.73.2 GB96%
Q6_K6.563.6 GB97%
Q8_08.54.5 GB100%
FP16168.1 GB100%
§ 01BENCHMARK SCORES
HumanEval80.5
MMLU-PRO52.8
MATH71.2
IFEval66.8
BBH46.5
GPQA24.6
MUSR16.8
BigCodeBench33.1
Arena Elo1465.0
§ 02RUN COMMAND

Run Phi-4-mini 3.8B locally with Ollama — needs 2.8 GB VRAM at Q4_K_M:

$ollama run phi4-mini:3.8b
§ 03COMPATIBLE GPUs
30 @ Q4_K_M
NVIDIA Tesla C2050
3 GB · 144 GB/s
NVIDIA Tesla M2050
3 GB · 148 GB/s
NVIDIA Tesla S2050
3 GB · 148 GB/s
NVIDIA GeForce GTX 670MX
3 GB · 67 GB/s
AMD Radeon HD 7950
3 GB · 240 GB/s
AMD Radeon HD 7950 Boost
3 GB · 240 GB/s
AMD Radeon HD 7950 Monica BIOS 1
3 GB · 240 GB/s
AMD Radeon HD 7950 Monica BIOS 2
3 GB · 240 GB/s
AMD Radeon HD 7970
3 GB · 264 GB/s
AMD Radeon HD 7970 GHz Edition
3 GB · 288 GB/s
AMD Radeon HD 7970 X2
3 GB · 264 GB/s
NVIDIA GeForce GTX 770M
3 GB · 96 GB/s
NVIDIA GeForce GTX 780
3 GB · 288 GB/s
NVIDIA GeForce GTX 780 Rev. 2
3 GB · 288 GB/s
NVIDIA GeForce GTX 780 Ti
3 GB · 337 GB/s