Gemma 4
Google's Apache 2.0 open models, from phone-sized E2B to a 31B flagship, with reasoning, vision and tool use.
In short. Gemma 4 is a family of free AI models from Google that you can download and run yourself. Small versions fit on a phone, mid-size ones run on a laptop or a gaming graphics card. They read text and images, think step by step and answer in more than 35 languages.
Summary
Gemma 4 is the latest generation of Google DeepMind's open models and the first to use the Apache 2.0 license, an open-source license approved by the Open Source Initiative, instead of the custom Gemma terms used before.
The family comes in five sizes. E2B and E4B are built for phones and small devices and understand text, images and audio. The 12B model is an encoder-free design that feeds images and audio straight into a single transformer. The 26B A4B is a mixture-of-experts model: it stores about 25 billion parameters but uses only 3.8 billion for each token, so it generates text much faster than its size suggests. The 31B dense model is the quality flagship.
Every size can reason step by step with a configurable thinking mode, call tools and read images. Context windows are 128K tokens for the two small models and 256K for the others. Google reports that the unquantized 26B and 31B models fit on a single 80 GB GPU, and 4-bit versions run on consumer graphics cards.
Good for
- Running a capable multimodal assistant on one GPU or a laptop
- On-device and offline apps, using the E2B and E4B sizes
- Products that need a permissive license without revenue limits
- Coding, reasoning and tool-calling agents on modest hardware
- Fine-tuning on your own data, especially the 31B dense model
Watch out for
- Benchmark scores on the model card come from Google and were not reproduced independently
- Only E2B, E4B and 12B accept audio input; the 26B and 31B models take text and images
- Long contexts need far more memory than the weights alone
- It writes text only and cannot generate images or speech
Performance claims come from the developer and have not been independently verified.
How much memory does it need?
Estimates for the E2B variant. A mixture-of-experts model must hold all of its experts in memory even though only some are used for each token, so total parameters decide the memory and active parameters decide the speed.
| Variants | Parameters | BF16 | FP8 / INT8 | 4-bit (Q4) | Fits on |
|---|---|---|---|---|---|
| E2B BF16 | 5.1B 2.3B active | 12 GB | 5.86 GB | 3.52 GB | Laptop or phone |
| E4B BF16 | 8B 4.5B active | 18 GB | 9.2 GB | 5.52 GB | Laptop or phone |
| 12B Unified BF16 | 11.9B | 27 GB | 14 GB | 8.25 GB | Laptop or phone |
| 26B A4B BF16 | 25.2B 3.8B active | 58 GB | 29 GB | 17 GB | Gaming GPU |
| 31B BF16 | 30.7B | 71 GB | 35 GB | 21 GB | Gaming GPU |
Estimates: weights plus about 15% overhead. Real usage grows with context length and depends on the software.
The license in plain words
One of the most widely used permissive licenses. You may use, change and sell products built on the model with only light conditions.
See also
- Gemma 4 vs Qwen3.8-27B: which open model should you run on one GPU?
- Best open-source LLMs in 2026
- Best open-source multimodal models in 2026 (images, video and audio)
- Best small open models you can run on a laptop
- Apertus 1.5: Switzerland's fully open model reads images and audio
- EmbeddingGemma 2 embeds text, images, video and audio under Apache 2.0
- Open models that listen: nine take audio, but only three answer in speech
- Raon-1: KRAFTON's Korean-first 28.8B model and a 3.1B on-device sibling
Frequently asked questions
- What is Gemma 4?
- Gemma 4 is a family of free AI models from Google that you can download and run yourself. Small versions fit on a phone, mid-size ones run on a laptop or a gaming graphics card. They read text and images, think step by step and answer in more than 35 languages.
- Can I use Gemma 4 commercially?
- Yes. Gemma 4 is released under Apache 2.0, a permissive license. You can use it commercially if you follow the standard conditions, such as keeping the license notice.
- Is Gemma 4 open source?
- Gemma 4 is released under Apache 2.0, a license approved by the Open Source Initiative. Training data and training code may not be public, so it is best described as open weights under an open-source license.
- Can I run Gemma 4 on my own computer?
- Yes. The smallest variant, E2B, needs about 3.52 GB of memory at 4-bit precision, which fits a modern laptop or a mid-range graphics card.
- When was Gemma 4 released?
- Gemma 4 was released by Google DeepMind on April 2, 2026.
- How long a context window does Gemma 4 have?
- Gemma 4 supports a context window of up to 256K tokens.
Official links
Sources
Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.
- Google blog: Gemma 4, byte for byte the most capable open models (2 Apr 2026) blog.google
- Google Open Source Blog: Gemma 4 and Apache 2.0 opensource.googleblog.com
- Hugging Face blog: Welcome Gemma 4 huggingface.co
- Gemma 4 31B model card huggingface.co
- Gemma release notes (Google AI for Developers) ai.google.dev
Release date and facts last verified November 8, 2026.
Related models
| Model | Released | Parameters | Context | License | Smallest variant needs |
|---|---|---|---|---|---|
| Qwen3.8-27B Alibaba Qwen | 27.8B | 256K | Apache 2.0 (Permissive) | 19 GB Gaming GPU | |
| gpt-oss OpenAI | 21B – 117B 5.1B active | 128K | Apache 2.0 (Permissive) | 15 GB Laptop or phone | |
| MiniCPM-V 4.7 OpenBMB | 1.55B – 35.2B 3B active | 256K | Apache 2.0 (Permissive) | 1.07 GB Laptop or phone | |
| Granite 4.2 IBM | 3.7B – 29.3B | 128K | Apache 2.0 (Permissive) | 2.55 GB Laptop or phone |