Latest Content
Gemma 3 4B IT icon

Gemma 3 4B IT

NVIDIA
Gemma 3 4B IT is an instruction-tuned multimodal transformer model built using research and technology from Gemini models. The model integrates text and vision processing, featuring a 4B parameter-class architecture with 34 transformer layers, a 2,560 hidden size, and a multimodal design combining a language backbone with a SigLIP-based vision encoder. It supports a 128K token context window with RoPE scaling for long-context understanding and can process images at 896×896 resolution encoded into 256 tokens.
TypeVision-Language Model
CapabilitiesText Generation, Instruction Following, Reasoning, Mathematical Reasoning+6 more
Links
Paper|Blog|HF Model Card
LicenseGemma

Inference Instructions

Deploy and run this model on NVIDIA B200 GPUs using the command below. Copy the command to get started with inference.

CONSOLE
docker run -it --rm 
 --runtime=nvidia 
 --gpus all 
 --ipc=host 
 --shm-size=64g 
 -p 8000:8000 
 -v ~/.cache/huggingface:/root/.cache/huggingface 
 -e HF_TOKEN='YOUR_HF_TOKEN' 
 -e LD_LIBRARY_PATH='/usr/local/nvidia/lib64:/usr/local/nvidia/lib:/usr/lib/x86_64-linux-gnu' 
 vllm/vllm-openai:v0.15.0-cu130 
 google/gemma-3-4b-it 
  --tensor-parallel-size 8 
 --max-model-len auto 
  --max-num-batched-tokens 65536 
  --gpu-memory-utilization 0.95 
 --max-num-seqs 1024 
 --disable-log-requests 
 --trust-remote-code

Model Benchmarks

Each model was tested with a fixed input size and total token volume while increasing concurrency to measure serving performance under load.

ITL vs Concurrency

Time to First Token

Throughput Scaling

Total Tokens/sec vs Avg TTFT

Vultr Cloud GPU

NVIDIA HGX B200

Deploy NVIDIA B200 on Vultr Cloud GPU

Deploy NVIDIA B200 instancesContact Sales