Hugging Face Overview
Gemma 3 27B Instruction-Tuned Model
This model is a 27 billion parameter instruction-tuned variant from the Gemma 3 family, developed by Google DeepMind. It is built upon the same research and technology as the Gemini models, offering open weights for both pre-trained and instruction-tuned versions. The model is multimodal, capable of processing both text and image inputs (normalized to 896x896 resolution, encoded to 256 tokens each) and generating text outputs.
Key Capabilities
- Multimodality: Processes text and image inputs, generating text outputs.
- Large Context Window: Supports a total input context of 128K tokens and an output context of 8192 tokens.
- Multilingual Support: Trained on data including content in over 140 languages.
- Diverse Training: Trained on 14 trillion tokens, encompassing web documents, code, mathematics, and images.
- Performance: Achieves strong results across various benchmarks, including 85.6 on HellaSwag (10-shot), 78.6 on MMLU (5-shot), and 65.6 on MBPP (3-shot).
Good For
- Content Creation: Generating creative text formats, marketing copy, and email drafts.
- Conversational AI: Powering chatbots and virtual assistants.
- Summarization: Creating concise summaries of documents and research papers.
- Image Understanding: Extracting, interpreting, and summarizing visual data.
- Research & Education: Serving as a foundation for VLM and NLP research, and supporting language learning tools.