Google Launches Multiple Open-Source Translation Models Following ChatGPT Translate

Google has continued its aggressive push in the artificial intelligence (AI) space in 2026. Following a series of initiatives including a partnership with Apple, the launch of new shopping tools, the introduction of Personal Intelligence in Gemini, and the integration of a chatbot into its Trends website, the company is now focusing on the open community. Its latest move comes with the release of TranslateGemma, a set of multilingual AI models designed for translation across a wide range of languages, supporting both text and image (input only) modalities.

In a recent blog post, Google announced three different variants of the TranslateGemma models. These AI models are available for download through Google’s Hugging Face listing and Kaggle, and can also be accessed via Vertex AI, the company’s cloud-based AI hub. Google has released the models under a permissive licence, enabling both academic and commercial use cases, which encourages broader adoption and experimentation by developers and enterprises alike.

The TranslateGemma models come in three sizes: 4B, 12B, and 27B parameters. The smallest, 4B model, is optimized for mobile and edge deployment, while the 12B variant targets consumer laptops, balancing performance and efficiency. The largest, 27B model, provides maximum translation fidelity and can run locally on a single Nvidia H100 GPU or TPU, making it suitable for high-demand environments and research applications.

With this release, Google is aiming to democratize AI-powered translation and provide the developer community with robust, versatile tools. By making TranslateGemma models open-source and easy to deploy across different devices, Google is reinforcing its commitment to accessible AI and strengthening its position in the multilingual AI ecosystem. The move also highlights the growing importance of translation technology in enabling global communication and cross-cultural collaboration.