Gemma 3 270M
Gemma 3 270M is the smallest member of Google's Gemma 3 family, a 270-million-parameter text-only model designed for on-device deployment and task-specific fine-tuning. It carries the Gemma license and Google's acceptable-use restrictions but is permitted for commercial use.
Not a chat model — treat it as a fine-tuning substrate. Gemma 3 270M is what you reach for when you want to ship a domain-specific SLM on a watch, hearing aid, or low-end Android device and you control the training data.
Overview
Gemma 3 270M is the smallest member of Google's Gemma 3 family, a 270-million-parameter text-only model designed for on-device deployment and task-specific fine-tuning. It carries the Gemma license and Google's acceptable-use restrictions but is permitted for commercial use.
Strengths
- True sub-300M footprint runs comfortably on a phone CPU
- 32K context window is exceptional at this size
- Excellent base for fine-tuning narrow tasks (intent, NER, routing)
- Backed by Google's full Gemma tooling (Keras, MediaPipe, Vertex)
Weaknesses
- Out-of-the-box chat quality is weak compared to Qwen3-0.6B
- Gemma license includes a prohibited-use policy that some legal teams dislike
- Gated repo requires HF acceptance even for download automation
- No instruction-tuned variant of this exact size, so requires fine-tuning
Quantization variants
Each quantization trades model quality for file size and VRAM. Q4_K_M is the most popular starting point.
| Quantization | File size | VRAM required |
|---|---|---|
| Q4_K_M | 0.2 GB | 1 GB |
Get the model
HuggingFace
Original weights
Source repository — direct quantization required.
Hardware that runs this
Cards with enough VRAM for at least one quantization of Gemma 3 270M.
Models worth comparing
Same parameter band, plus what's one tier above and below — so you can decide what actually fits your hardware.
Frequently asked
What's the minimum VRAM to run Gemma 3 270M?
Can I use Gemma 3 270M commercially?
What's the context length of Gemma 3 270M?
Source: huggingface.co/google/gemma-3-270m
Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify model claims.
Related — keep moving
Verify Gemma 3 270M runs on your specific hardware before committing money.