# About What Can I Run AI? | What Can I Run AI?

> Who runs What Can I Run AI? and how the data is made.

Source: https://whatcanirunai.vercel.app/about/

# About What Can I Run AI?

What Can I Run AI? is a PCPartPicker for AI workloads. It answers one question well: which open language models will run on a given GPU, Mac or mini PC, and what you should buy to run the model you want.

## What we do

We track 656 popular open-weight models and 82 GPUs, Macs and mini PCs. For every pair we estimate the memory the model needs at Q4\_K\_M, Q8\_0 and FP16 quantization, whether it fits in the memory the device can actually use, and a calibrated range for decode speed in tokens per second. The [AI PC planner](https://whatcanirunai.vercel.app/planner/) adds CPU, RAM, power supply and case checks so a build is complete and safe, and the [ready-made builds](https://whatcanirunai.vercel.app/builds/) turn a goal like "run a 70B model" into an exact buy list.

## Where the numbers come from

Parameter counts come from the Hugging Face API for each model. GPU memory, bandwidth and board power come from NVIDIA, AMD, Apple and Intel spec pages. Prices are manufacturer list prices (launch MSRP for discontinued cards) with a dated source, not live street prices. Every formula is published on [How we calculate](https://whatcanirunai.vercel.app/method/), including the efficiency bands we use for speed and how they were calibrated against llama.cpp community benchmarks.

## How we make money

Some buy buttons are affiliate links, which pay us a small commission at no cost to you. Rankings never depend on commissions. Agents and developers can use the [free JSON API](https://whatcanirunai.vercel.app/developers/).

## Contact

Questions, corrections or a model you want added: see [Contact](https://whatcanirunai.vercel.app/contact/).

---
Agent guide: https://whatcanirunai.vercel.app/llms.txt · JSON API: https://whatcanirunai.vercel.app/developers/