TranslateGemma — local AI model by Google DeepMind
Written by Jakub Rusinowski · Last updated
Google's open translation models, fine-tuned from Gemma 3 in 4B, 12B and 27B sizes and evaluated on 55 languages. Google reports the 12B beating the general Gemma 3 27B on translation quality, and the models keep Gemma 3's ability to translate text inside images. The weights use the Gemma Terms of Use (not Apache 2.0) and the Hugging Face repositories ask you to accept that licence before downloading. They are translation specialists, so they never appear in this site's "best model" rankings.
Variants
The smallest TranslateGemma variant needs about 3 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| TranslateGemma 4B → 4B | ~3.4 GB | 2,048 (input) | ollama run translategemma:4b | |
| TranslateGemma 12B → 12B | ~8.2 GB | 2,048 (input) | ollama run translategemma:12b | |
| TranslateGemma 27B → 27B | ~17.3 GB | 2,048 (input) | ollama run translategemma:27b |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run TranslateGemma locally
Install Ollama, then pull the tag.
ollama run translategemma:4bPick a size above for its own VRAM figure, speed estimate and install command.
Licence
Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.
Applies to: TranslateGemma 4B, TranslateGemma 12B, TranslateGemma 27BRecommended GPU
The cheapest catalogued GPU that runs TranslateGemma locally (min 3 GB VRAM) is the Intel Arc B570 (10 GB).
TranslateGemma — frequently asked questions
How much VRAM does TranslateGemma need?
TranslateGemma needs about 3 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: TranslateGemma 4B (3 GB, Q4_K_M); TranslateGemma 12B (8 GB, Q4_K_M); TranslateGemma 27B (17 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run TranslateGemma on an RTX 4090 (24 GB)?
Yes — TranslateGemma runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for TranslateGemma?
Q4_K_M is the best balance of quality and VRAM for TranslateGemma in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run TranslateGemma with Ollama?
Install Ollama, then run: ollama run translategemma:4b. This downloads TranslateGemma and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.