Welcome to the Gen AI Model Reference repository — a resource for AI Engineers looking to understand and compare real-world hardware requirements for running generative AI models.
🔗 Live site: imraf.github.io/ai-model-reference
This repository powers the AI Model Reference website, which serves as a curated reference for AI models. It includes:
- Static web pages (for GitHub Pages)
- CSV data files containing model hardware requirements
➤ These are also accessible from the website interface.
The goal is to reflect real, tested hardware requirements — not theoretical or estimated specs.
Please ensure any data you submit is verified, meaning it’s tested on actual hardware by monitoring VRAM usage during model operation.
There are two ways to contribute:
- Open contributor?
- Clone this repository
- Add your new entries to
data/vram.csv - Make sure your data is validated using real hardware
- Contact me:
Feel free to reach with new data points and suggestions. I'm active on:
- MDLI group
- LinkedIn (From time to time)
- The initial batch of data was recorded from model cards on HuggingFace, measured by myself and scraped from the project Can I Run This LLM?, which (at the time of writing) is no longer available.