Gemma Family Expands with Models Tailored for Developers and Researchers

Gemma, a family of lightweight open models announced in February, has expanded with two new variants, CodeGemma and RecurrentGemma, designed to enhance machine learning capabilities in code completion and research experimentation. CodeGemma provides developers with advanced code completion, generation, and chat functions, available in multiple pretrained and instruction-tuned variants, supporting languages like Python and JavaScript, and optimized for both local and cloud environments. RecurrentGemma, distinct from traditional Transformer models, utilizes recurrent neural networks for efficient memory usage and faster inference at higher batch sizes, making it suitable for devices with limited memory. Both models are built on the principles of open availability and responsible AI, compatible with various software and hardware platforms including JAX, PyTorch, and NVIDIA GPUs. Alongside these models, Gemma 1.1 has been released with performance improvements and updated terms based on user feedback. The models are accessible on platforms like Kaggle, Hugging Face, and Vertex AI Model Garden, inviting developers to explore and innovate with these advanced tools.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.