NelsaHost
search Login
smart_toy
Phi Large Language Models

Phi-3 Mini 3.8B (3.80B)

Parameters
3.80B
VRAM (FP16)
7.6GB
VRAM (INT4)
1.9GB
Context
128000
check_circle View Compatible GPUs open_in_new View on HuggingFace

tune Quantization Options

Quantization VRAM Required Min GPU
FP16 (Half Precision) 7.6GB RTX 3070 / 4060
INT8 (8-bit Integer) 3.8GB RTX 3070 / 4060
Q4_K_M (GGUF 4-bit) 1.9GB RTX 3070 / 4060
q3_k_m 1.5GB RTX 3070 / 4060

memory Compatible GPUs

check_circle

RTX 4090

24.0GB VRAM

Perfect
check_circle

RTX 4090

24.0GB VRAM

Perfect
check_circle

RTX 4090

24.0GB VRAM

Perfect
check_circle

RTX 4090

24.0GB VRAM

Perfect
check_circle

RTX 3090 Ti

24.0GB VRAM

Perfect
check_circle

RTX 3090 Ti

24.0GB VRAM

Perfect
View All Compatible GPUs

Model Details

Family Phi
Category Large Language Models
Parameters 3.80B
Context Length 128000

Similar Models

Phi-3 Medium 14B 14.00B Phi-3 Small 7B 7.00B