Google has released the artificial intelligence model LLM Gemma 2, which was announced at the recent Google I/O event. It comes in two sizes to choose from, 9B and 27B, emphasizing increased intelligence, improved speed, and safe response training.
Testing reports of Gemma 2 on Chatbot Arena reveal that the 27B variant outperforms the Gemini 1.0 Pro (although the test sample size is limited) while the 9B model surpasses Claude 3 Haiku with an impressive average score.
Google emphasizes the importance of safe model training with Gemma, filtering out inappropriate data for training and enhancing responses to ensure appropriate behavior in various scenarios.
This model is now available for download on Hugging Face and can be used with various tools such as Gemma.cpp, vLLM, Keras 3.0, Llama.cpp, Ollama, and NVIDIA NIM for fine-tuning on Hugging Face. Users can also utilize it on Google AI Studio immediately and will soon be able to access it on Vertex AI for cloud usage.
TLDR: Google introduces LLM Gemma 2, offering two sizes for enhanced intelligence and response training while emphasizing safety in model training and providing versatile usage options on platforms like Hugging Face, Google AI Studio, and Vertex AI.
Leave a Comment