Gemini vs Gemma
People think Gemma is just a small Gemini, and it is not. Gemini runs on Google's servers, and you reach it online. Gemma is its own family of open models. You download it and run it on your own phone or laptop. It is smaller, so it costs nothing per token but does less on hard tasks.
Gemini
Google's AI models. They take in text, pictures, sound and video.
Gemini is Google DeepMind's family of AI models, first shown in December 2023. It was built from the start to take in text, images, audio and video together. It comes in sizes such as Flash, which is fast, and Pro, which is stronger. Gemini runs across Google's products, from Search to Gmail. Gemini is also the name of Google's chat app, which used to be called Bard. A smaller sister family, Gemma, is released with open weights.
Gemma
Open models from Google. You get them and run them on your own.
Gemma is Google DeepMind's family of open models. You can download them and run them on your own phone, laptop or server. They are smaller than Gemini, Google's main models. Gemma 4 came out in April 2026 in five sizes, from E2B for phones up to 31B. All of them read text and pictures, and the small sizes can also hear sound.
In one line
- Gemini
- Google's AI models. They take in text, pictures, sound and video.
- Gemma
- Open models from Google. You get them and run them on your own.