Google has unveiled Gemma, a new family of AI language models that are free, open-weights models built on technology, allowing them to run locally on a desktop or laptop. Developed by Google DeepMind and other Google AI teams, the Gemma models are available in two sizes: Gemma 2B (2 billion parameters) and Gemma 7B (7 billion parameters). Google claims that the 7B model outperforms Meta’s Llama 2 7B and 13B models on several benchmarks for math, Python code generation, general knowledge, and commonsense reasoning tasks. In addition, Google has also released a “Responsible Generative AI Toolkit,” which aims to provide guidance and tools for developing “safe and responsible” AI applications.
This release is Google’s first significant open large language model (LLM) since OpenAI’s ChatGPT, which started a frenzy for AI chatbots in 2022. Although Gemma is not open source, it feels like a conspicuous play to match Meta, which has been releasing open-weights models since last year. Google’s move intends to attract more developers to its Vertex AI cloud platform and seems to stand in opposition to AI models like OpenAI’s GPT-4 Turbo, which is available only through the ChatGPT application and a cloud API and cannot be run locally.
In essence, the introduction of Gemma models is seen as an essential step in ensuring transparency and privacy in chatbots. Google’s offering of a new family of AI models provides another layer of choice and opportunity for developers, and it will be interesting to see how these models will impact the AI landscape in the coming years.