Compare LLM Models

Select 2-5 models to compare side by side.

Select models to compare 0 / 5 selected

Top downloads

🤖 DeepSeek-V4-Pro-0813 deepseek-ai 🤖 GLM-5.3 zai-org 🤖 Kimi-K3 moonshotai

Select 2-5 models

Attribute	🤖 medgemma-4b-it google
Company	google
Release Date	2025-05-20
Parameters (B)	4
Architecture	Decoder-only Transformer
Context Window	128000
Input Modalities	text, image
Open Source	Yes
License	other
Score	0
VRAM (GB)	8
Compute	Single NVIDIA GPU with ≥8 GB VRAM for bfloat16 inference (~8.6 GB weights); 10–12 GB recommended for multimodal image+text workloads. Runs via Hugging Face Transformers (≥4.50); quantized GGUF variants can run on 6–8 GB.
Benchmark: MMLU-Pro	39.1
Benchmark: MMMU	47.3
Pricing (Input)	—
Pricing (Output)	—