Downloadtranslategemma-4b-it is a lightweight, state-of-the-art open translation model from Google, built on the Gemma 3 family and optimized for high-quality multilingual translation across 55 languages. It supports both text-to-text translation and image-to-text extraction with translation, enabling workflows such as OCR-style translation of signs, documents, and screenshots. With a compact ~5B parameter footprint and BF16 support, the model is designed to run efficiently on laptops, desktops, and private cloud infrastructure, making advanced translation accessible without heavy hardware requirements. TranslateGemma uses a structured chat template that enforces explicit source and target language codes, ensuring consistent, deterministic behavior and reducing ambiguity in multilingual pipelines. It integrates seamlessly with Hugging Face Transformers through pipelines or direct model initialization, supporting GPU acceleration and scalable deployment.
Features
- Multilingual translation across 55 supported languages
- Image text extraction and translation in a single pipeline
- Lightweight ~5B parameter footprint for local or private deployment
- Structured chat template with explicit source and target language control
- Hugging Face Transformers compatibility with pipeline and direct loading
- GPU acceleration with BF16 for efficient inference
- Strong benchmark performance on WMT and multimodal datasets
- Designed with safety evaluation, bias mitigation, and responsible use in mind
Project Samples
