Taking a clear step toward open and flexible AI translation, Google has introduced a new set of multilingual models designed for developers who want more control over how translation systems are built and deployed.
The company announced TranslateGemma, a family of open translation models built on the Gemma 3 architecture. The models support translation across 55 languages and are positioned as an open alternative to closed, cloud-based translation systems, including ChatGPT’s translation capabilities.
TranslateGemma is aimed at researchers and developers looking for efficient, locally deployable models rather than proprietary services that rely entirely on external servers. Google said the focus is on openness, efficiency, and flexibility across different environments.
The models are released in 3 sizes — 4B, 12B, and 27B parameters — covering use cases from mobile and edge devices to large-scale cloud deployments. Google said the models are trained using a 2-stage process, combining supervised fine-tuning with reinforcement learning, based on high-quality human and synthetic translation data.
According to the company, TranslateGemma reduces translation error rates across high-, mid-, and low-resource languages when compared with baseline Gemma models, while using fewer parameters.
A key difference from ChatGPT’s translation workflow is access. ChatGPT’s translation feature operates as a closed system, while TranslateGemma offers open weights that developers can download, inspect, and fine-tune. This allows the models to run locally on devices, private servers, or custom hardware, without sending data to external servers.
This approach is particularly relevant for enterprises and researchers working with sensitive data or in low-connectivity environments, where cloud-only translation tools may not be suitable.
TranslateGemma supports 55 evaluated language pairs and has been trained on nearly 500 additional language pairs for future experimentation. Google also noted that the models retain multimodal capabilities from Gemma 3, allowing text translation within images without requiring separate multimodal training.
This makes it possible to extend translation beyond plain text to image-based content and documents.
TranslateGemma models are available through platforms including Kaggle, Hugging Face, Google Colab, and Vertex AI. Google has also released a technical report outlining training methods, benchmarks, and supported languages.
With this launch, Google is positioning open translation models as a practical alternative to proprietary AI translation tools, giving developers greater control over deployment, customization, and data handling.
Also read: Viksit Workforce for a Viksit Bharat
Do Follow: The Mainstream formerly known as CIO News LinkedIn Account | The Mainstream formerly known as CIO News Facebook | The Mainstream formerly known as CIO News Youtube | The Mainstream formerly known as CIO News Twitter
About us:
The Mainstream is a premier platform delivering the latest updates and informed perspectives across the technology business and cyber landscape. Built on research-driven, thought leadership and original intellectual property, The Mainstream also curates summits & conferences that convene decision makers to explore how technology reshapes industries and leadership. With a growing presence in India and globally across the Middle East, Africa, ASEAN, the USA, the UK and Australia, The Mainstream carries a vision to bring the latest happenings and insights to 8.2 billion people and to place technology at the centre of conversation for leaders navigating the future.
