Google’s new Gemma AI models have already hit a historic milestone, reaching over 150 million downloads since their release in February 2024. Omar Sanseviero, a developer relations engineer at Google DeepMind, shared this breakthrough on X. The rapidity of this uptake reflects a growing hunger for AI models that can soon compete with the existing players in the industry. Just as one case in point, Meta’s Llama has already been downloaded over 1.2 billion times as of late April.
Launched to compete against other much-hyped “open” model families, Gemma has quickly made waves in the regions for its disruptive potential. Developers have built more than 70,000 different versions of Gemma on the AI development platform Hugging Face. This increase is an important indicator of a vibrant and engaged community that is actively pushing the model in new and exciting directions.
Google’s development center for Gemma is situated in Arnulfpark, Munich, where more than 2,500 employees contribute to the company’s AI initiatives across Germany. The recent multimodal releases of Gemma have expanded these multimodal functionalities, enabling users to interact with images and text together. In addition to greater accuracy, these updates offer support for over 100 languages, deepening their reach and user experience in various multicultural markets.
Even with all of these achievements, Gemma absolutely deserves criticism in terms of its licensing terms. Other developers express concern over the proprietary and non-standard licensing agreements. Collectively, they view these as the type of factors that render Gemma’s commercial use a poor bet. This has led to great alarm about whether it can continue to grow at such an unsustainable pace. By comparison, competitors such as Llama offer a much clearer and simpler licensing based.
