google/gemma-3-270m
Gemma 3 270M is a lightweight, multimodal open model from Google DeepMind, part of the Gemma family, built using the same research and technology as Gemini models. This 270 million parameter model handles both text and image inputs, generating text outputs, and supports a 32K token context window. It is designed for text generation and image understanding tasks like question answering, summarization, and reasoning, making it suitable for deployment in resource-limited environments.
Loading preview...
Gemma 3 270M: A Lightweight Multimodal AI Model
Gemma 3 270M, developed by Google DeepMind, is a compact yet powerful open model from the Gemma family, leveraging the same foundational research as the Gemini models. This 270 million parameter variant is designed for efficiency and accessibility, making advanced AI capabilities available for deployment in environments with limited computational resources, such as laptops or edge devices.
Key Capabilities
- Multimodal Input: Processes both text strings and images (normalized to 896 x 896 resolution, encoded to 256 tokens each for larger Gemma 3 models, though the 270M model's image input specifics are not detailed beyond general multimodal capability).
- Text Generation: Generates text outputs in response to diverse inputs, including answers to questions, summaries of documents, and analysis of image content.
- Context Window: Supports a total input and output context of 32K tokens, enabling processing of moderately long sequences.
- Multilingual Support: Trained on data including content in over 140 languages, indicating multilingual capabilities.
- Broad Task Suitability: Well-suited for a variety of tasks such as question answering, summarization, and reasoning.
Training and Performance
The 270M model was trained on 6 trillion tokens, with a knowledge cutoff date of August 2024. The training dataset included web documents, code, mathematics, and images, ensuring exposure to a wide range of linguistic styles and data formats. Rigorous data cleaning and filtering, including CSAM and sensitive data filtering, were applied. Benchmarks for the Gemma 3 270M model show scores such as 40.9 on HellaSwag (10-shot PT) and 67.7 on PIQA (0-shot PT), demonstrating its foundational capabilities for its size.
Intended Use Cases
- Content Creation: Generating creative text formats, marketing copy, or email drafts.
- Conversational AI: Powering chatbots and virtual assistants.
- Text Summarization: Creating concise summaries of documents.
- Image Data Extraction: Interpreting and summarizing visual data for text communications.
- Research and Education: Serving as a foundation for VLM and NLP research, language learning tools, and knowledge exploration.