Google - Dense - verified
Run Gemma 4 12B locally
Gemma 4 12B is a 12B dense model from Google with a 262K-token context window, released under Apache 2.0 (commercial use is allowed); the smallest listed download is 6.7 GB, the 4bit MLX build from mlx-community/gemma-4-12B-it-4bit.
Official weights: google/gemma-4-12B-it - modalities: text, image, audio, video
- Total params
- 12B
- Architecture
- Dense
- Context
- 262K
- Smallest download
- 6.7 GB
Fits in
Smallest download at most 85% of memoryDownloads
The official weights and every quantization listed for this model, with file sizes from Hugging Face.
| Download | Format | Size | Repo |
|---|---|---|---|
Official weights | BF16 | 23.9 GB | google/gemma-4-12B-it |
Q4_0 (QAT) | GGUF | 7.0 GB | google/gemma-4-12B-it-qat-q4_0-gguf |
Q4_K_M | GGUF | 7.1 GB | unsloth/gemma-4-12b-it-GGUF |
Q8_0 | GGUF | 12.7 GB | unsloth/gemma-4-12b-it-GGUF |
4bitSmallest listed | MLX | 6.7 GB | mlx-community/gemma-4-12B-it-4bit |
GB is decimal: 1 GB = 10^9 bytes. Add memory for the context and the operating system on top of the file size.
Run it
Copied verbatim from the page linked under each command. Server commands from model cards often assume a multi-GPU node; check the flags before you run them.
Transformers
pip install -U transformers torch accelerateNot sure which runtime to use? Compare local LLM runtimes by platform, API and features.
License
- License
- Apache 2.0
apache-2.0 - Commercial use
- Commercial use: YesCommercial use is allowed.
- Gated download
- No. The files download directly.
Worth knowing
Encoder-free unified model that projects image patches and audio waveforms straight into the decoder, released two months after the rest of the Gemma 4 family.
Released (per the announcement).
FAQ
How big is the Gemma 4 12B download?
The official BF16 weights in google/gemma-4-12B-it are 23.9 GB. The smallest listed download is 6.7 GB, the 4bit MLX build from mlx-community/gemma-4-12B-it-4bit. Plan for more memory than the file size: the context (KV cache) and the operating system need room too.
Can I use Gemma 4 12B commercially?
Gemma 4 12B is released under Apache 2.0, and commercial use is allowed. The license text is at https://ai.google.dev/gemma/docs/gemma_4_license.
How much memory do I need for Gemma 4 12B?
By this directory's rule of thumb it fits in 8 GB and up: the smallest listed download, 6.7 GB, is at most 85% of 8 GB, which leaves room for the context and the operating system.
How do I run Gemma 4 12B with Ollama?
Run "ollama run gemma4:12b", as shown at https://ollama.com/library/gemma4. The Ollama library lists the gemma4:12b tag at 7.7GB - 8.0GB.
What is the context length of Gemma 4 12B?
262,144 tokens. A longer context needs more memory for the KV cache, so you can run with a shorter one when memory is tight.
Sources
Last verified from the pages below.
- HF API - huggingface.co
- HF file tree - huggingface.co
- Model card (family table) - huggingface.co
- Ollama gemma4 - ollama.com
- Google blog: Introducing Gemma 4 12B - blog.google
Newsletter
New open weights, every week
One email per week. New model releases, open-weight drops and what they mean for what you build.
Keep exploring
Best local AI models
Our curated picks by hardware tier, with benchmarks and the reasoning.
Model API pricing and specs
First-party prices per 1M tokens when you would rather call an API.
AI benchmark leaderboards
Sourced scores, for when you need to compare quality.
Setup guides
Point your coding tools at the model you picked.