Are Open Source Models Actually Free?

A look at the hidden costs behind open source AI models — compute, infrastructure, fine-tuning, and operational overhead that don't show up in the license.

Open source AI models are everywhere — Llama, Mistral, Qwen, Falcon, Gemma. The pitch is simple: free weights, full control, no API bills. But is that actually true?

In this video I break down the real costs of running open source models yourself — from GPU compute and cloud infrastructure to fine-tuning, serving, and maintenance — and compare that honestly against just paying for a hosted API.

The answer is more nuanced than the open source community usually admits.