AI Hardware Fit is an open-source, browser-based calculator for a question that is surprisingly time-consuming to answer: Can my GPU run this AI model?

Choose a GPU to see compatible models, quantization, VRAM, estimated speed, and ready-to-run Ollama or llama.cpp commands. Or choose a model to compare suitable GPUs by price, speed, and power. It can also estimate a server configuration and approximate cost for an AI service from user count, concurrency, and SLA targets.

It covers 151 GPU presets and 332 AI models across NVIDIA, AMD, Intel, and Apple Silicon, including LLMs, embeddings, rerankers, OCR/VLM, image and video generation, STT, and TTS. Results clearly label whether they come from measured data, measurement-calibrated estimates, related measurements, or calculated estimates.

No account or backend is required, and inputs stay in the browser. The formulas and data are open source under the MIT license. I am looking for real-world measurements and feedback to improve the recommendations.