AI · AI model
TranslateGemma (google/translategemma-4b-it)
Open translation model by Google DeepMind lightweight enough for on-device use.
TranslateGemma is an open language model variant from Google DeepMind optimized for text translation across dozens of languages. At 4B parameters, it is lightweight enough to be deployed on local hardware and mobile devices using frameworks like Apple MLX.
What it's for
- Run offline, on-device translation in mobile applications
- Translate text across 50+ languages without sending data to third-party APIs
- Deploy a dedicated local translation model on Apple silicon hardware
Features
- 4B parameter instruction-tuned model size
- Support for translation across 50+ languages
- Compatible with Hugging Face Transformers and Apple MLX
- Runs locally on edge and consumer hardware like iPhones and Macs
Pricing
Open source. Free under Gemma license terms
Details
- Type
- AI model
- Category
- AI
- Topic
- Translation Models
- Open source
- Yes
- Platforms
- Code library, Self-hosted, iOS, macOS, Linux
- For
- Mobile and AI developers building offline or private translation features.
- Alternative to
- Google Cloud Translation API, DeepL API, NLLB-200
- Tags
- translation, gemma, deepmind, on device, mlx, nlp
Why it was saved
- @adrgrondin: Named in the tweet as the model powering on-device translation on iPhone; relevant for anyone building local AI translation features.
Related
Offline, private AI translation app running directly on your iPhone.
Private, in-browser machine translation running Google's TranslateGemma model via WebGPU
AI avatar video generation and translation platform driven by text prompts
Generative AI video model producing cinematic footage with native audio
Dynamic video inspection for Gemini models that cuts token usage and cost
AI speech-to-text platform for multilingual transcription, translation, and video subtitling
Updated