Summary
TLDR: Google AI released Gemma, a new family of open-source language models that directly competes with Meta’s LLaMa and Mistral. Gemma is available in two versions, optimized for efficiency and better results. It has quickly climbed the ranks on the Hugging Face leaderboard and outperforms Meta’s LlaMa-2 in various benchmarks. Developers can test Gemma for free on Kaggle and Google Colab, and apply for Google Cloud credits for research projects.
Key Points
1. Google AI introduced Gemma, a new family of open-source language models, to compete with leading LLMs from Meta and OpenAI.
2. Gemma is available in two versions – Gemma 2B and Gemma 7B, with different parameter sizes and performance capabilities.
3. Google aims to democratize AI technology by releasing Gemma as an open-source model and providing resources for developers to test and use it effectively.