October 18, 2024: Compact AI Models for Multimodal Tasks Unveiled - H2O.ai has launched two new small language models, Mississippi 2B and Mississippi 0.8B, for multimodal tasks like text extraction from images. Mississippi 2B features 2.1 billion parameters, allowing it to generate detailed image descriptions and extract data for applications like sales logs. The scaled-down Mississippi 0.8B, with 800,000 parameters, excels in optical character recognition, outperforming comparable models. Both models operate efficiently on limited processing devices, offering fast response times for latency-sensitive tasks. Available on Hugging Face under an open-source license, these models are optimized for high-performance yet cost-effective AI deployment.
