Gemma 3 27B Instruction-Tuned INT4
| Property | Value |
|---|---|
| Original Model | Google Gemma 3 27B |
| Quantization | INT4 |
| Context Window | 128K tokens |
| Training Tokens | 14 trillion |
| License | Available on Terms of Use |
What is gemma-3-27b-it-int4-awq?
This model is a highly efficient INT4-quantized version of Google's Gemma 3 27B instruction-tuned model, converted to HF+AWQ format for easier deployment. It maintains the powerful capabilities of the original model while significantly reducing the memory footprint through INT4 quantization. The model supports both text and image inputs, making it a versatile option for multimodal applications.
Implementation Details
The model leverages quantization-aware training (QAT) for INT4 precision, converted from the original Flax checkpoint. It retains the architecture's 128K context window and supports over 140 languages. The implementation allows for efficient deployment on resource-constrained environments while maintaining high performance across various tasks.
- Multimodal capabilities with support for text and image inputs (896x896 resolution)
- Output generation up to 8192 tokens
- Optimized for deployment on consumer hardware
- Converted to HF+AWQ format for broader compatibility
Core Capabilities
- Strong performance in reasoning and factuality tasks (85.6% on HellaSwag)
- Robust STEM and code generation capabilities (82.6% on GSM8K)
- Multilingual support with strong performance (74.3% on MGSM)
- Advanced vision-language understanding (85.6% on DocVQA)
Frequently Asked Questions
Q: What makes this model unique?
The model combines the powerful capabilities of Gemma 3 with efficient INT4 quantization, making it possible to run a state-of-the-art multimodal model on consumer hardware while maintaining high performance across various benchmarks.
Q: What are the recommended use cases?
The model excels in content creation, chatbots, text summarization, image analysis, research applications, and educational tools. It's particularly well-suited for deployments where resource efficiency is crucial while maintaining high-quality outputs.
