Estimate GPU memory requirements for running AI models locally. Accurate calculations based on real transformer architecture.
Precise formulas based on real transformer architecture. Consider layers, hidden size, attention heads, and KV heads.
Support for Llama 3.2/3.1/3/2, Qwen 3.5/3/2.5, Mistral, Mixtral, Gemma, Phi, Command R, and more.
Search and load model configurations directly from HuggingFace. Support for thousands of models.
Consider 60-85% memory reduction with Flash Attention enabled. Optimized for modern hardware.
Automatic language detection. Available in English and Spanish, with more languages coming soon.
Automatic dark mode that follows your system preferences. Easy on the eyes, day or night.
Download the app and start calculating VRAM requirements for your favorite LLMs.