Developers who want capable open models to run and fine-tune themselves get four sizes under Apache 2.0, free even through the API.
Kenning's take
What it does
Gemma 4 is Google's open-weight model family released under the Apache 2.0 licence, built from the same research as Gemini 3. It comes in four sizes, from small models for phones to a 31B dense model, and is aimed at reasoning and agent tasks. The weights can be downloaded from Hugging Face, Kaggle or Ollama, and it is also free to call in the Gemini API.
- Four sizes: E2B, E4B, 26B MoE, 31B dense
- Apache 2.0 licence
- Runs on phones, laptops and workstations
- Multi-step reasoning and agent workflows
- Day-one support in Ollama, llama.cpp, MLX and vLLM
Who it suits. Developers and researchers who want capable models they can run and fine-tune themselves.
Key facts
- Category
- AI models
- Made by
- Based in
- United States
- Released
- 2 April 2026
- How to get it
- Download weights from Hugging Face, Kaggle or Ollama; free in the Gemini API
- Works on
- Local, API, Cloud
- In the UK
- Yes
- Free plan
- Yes, open weights and free Gemini API use
- API price
- Free of charge in the Gemini API (no paid tier listed)
FAQ
Is Gemma 4 free?
Yes. The weights are open under Apache 2.0 and Google lists Gemma 4 as free in the Gemini API.
What sizes does Gemma 4 come in?
Effective 2B, Effective 4B, a 26B mixture-of-experts model and a 31B dense model.
When was Gemma 4 released?
2 April 2026.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.