On April 2nd, Google released Gemma 4 — a set of AI models you can download, run on your own hardware, and use however you like, including commercially. No monthly fees. No permission slips. No fine print about how many users you can have.
The family comes in different sizes. The smallest fits on an Android phone. The largest sits comfortably on a decent workstation. The top one is currently the third best open AI model in the world — while being a fraction of the size of its competitors.
Most AI tools you use today are rented. You pay monthly, your data goes somewhere you can't see, and the moment the pricing changes, you're stuck.
With Gemma 4, you could run an AI that understands text, images, audio, and over 140 languages — entirely inside your own systems. A law firm keeping client documents private. A restaurant chain summarising customer feedback without sending it to a third-party server. A design agency building a client tool without worrying about API costs scaling out of control.
It's not that cloud AI is bad. It's that having a real alternative changes your negotiating position entirely.
Open-weight model — An AI whose inner workings are published for anyone to download and use. Like getting the actual recipe instead of just the dish.
Apache 2.0 license — A legal agreement that says: take this, modify it, sell with it, do what you want. No strings attached.
Context window — How much information an AI can hold in its head at once. A bigger window means it can read longer documents without forgetting the beginning.
Fine-tuning — Teaching an AI to specialise in your world — your industry, your tone, your data — without building one from scratch.
If you've been curious about running AI inside your own business without depending on subscriptions, this is a good moment to ask someone you trust: could we try this with something small?