H2O.ai Launches New Multimodal Foundation Models to Undertake Document AI Use Cases

Highlights:

  • H2OVL Mississippi 0.8B Model Surpasses Leading Small Vision Language Models (SVLMs) and Impressively Outperforms Larger State-of-the-Art Vision Language Models (VLMs) in OCR Benchmarks for Text Recognition
  • H2OVL Mississippi 2B Rivals State-of-the-Art SLMs on Single Image Benchmarks
  • New powerful OCR model powering Enterprise h2oGPTe Agentic RAG platform

H2OVL Mississippi 0.8B is the smallest, most cost efficient model delivering the best OCR for Text Recognition compared to other SLMs (Graphic: Business Wire)

MOUNTAIN VIEW, Calif.--()--H2O.ai, the leader in open-source Generative AI and most accurate Predictive AI platforms, today announced H2OVL Mississippi 2B and 0.8B, two powerful new multimodal foundation models designed specifically for OCR and Document AI use cases. Compact yet highly efficient, the H2OVL Mississippi foundation models represent a significant advancement in AI, delivering unmatched performance for vision and OCR tasks in enterprise environments.

Available now on Hugging Face, H2OVL Mississippi 2B and 0.8B offer enterprises an economical solution with efficiency and accuracy for real-time document analysis and image recognition.

Open Weight H2OVL Mississippi Vision and OCR: Free Access

H2O.ai’s decision to release H2OVL open weight model series has sparked significant interest within the AI community. By making the model freely accessible on Hugging Face, developers, researchers, and enterprises can now modify, fine-tune, and adapt H2OVL Mississippi models to fit their specific OCR and Document AI needs.

H2OVL Mississippi 2B builds on the legacy of H2O Danube2 with a robust 2.1 billion parameter model optimized for lightweight deployment and specialized multimodal architecture that blends language and computer vision to meet the growing demand for more economical multimodal OCR. Pre-trained on 5.3 million conversation pairs and fine-tuned with an additional 12 million pairs, H2OVL Mississippi 2B excels at handling diverse image resolutions, ranging from 448px to 4K.

Built on the Danube3 0.5B, H2OVL Mississippi 0.8B model—pre-trained on 11 million conversation pairs and fine-tuned with an additional 8 million—surpassed all comparable SLMs in the market on OCR benchmarks, delivering unmatched performance on text recognition.

“We’ve designed H2OVL Mississippi models to be a high-performance yet cost-effective solution, bringing AI-powered OCR, visual understanding, and Document AI to businesses,” said Sri Ambati, CEO and Founder of H2O.ai. “By blending state-of-the-art multimodal AI with extreme efficiency, H2OVL Mississippi delivers precise, scalable Document AI solutions across a range of industries.”

Key Features of H2OVL Mississippi 2B and 0.8B

  • Lightweight Model: 2B and 0.8B parameters optimized for efficient deployment, enabling powerful AI performance with minimal resource consumption.
  • Multimodal Mastery: Seamlessly handles OCR and Document AI tasks across varied resolutions, providing versatile vision-language capabilities.
  • Tailored Training: Multi-stage training with fine-tuning layers for highly customized application performance.
  • Real-Time Efficiency: Delivers real-time processing with minimal latency, making it ideal for industries such as banking, financial services, telco, manufacturing, healthcare, insurance, and the public sector where accurate document processing is crucial.

Now Available on Hugging Face

H2OVL Mississippi 2B https://huggingface.co/h2oai/h2ovl-mississippi-2b
H2OVL Mississippi 0.8B https://huggingface.co/h2oai/h2ovl-mississippi-800m

H2OVL Mississippi 2B is now available for developers, enterprises, and researchers on Hugging Face, making it easier than ever to integrate powerful, lightweight AI into workflows that require top-tier OCR and document understanding. It is also integrated into Enterprise h2oGPTe https://h2o.ai/platform/enterprise-h2ogpte/ for immediate value-add for existing RAG powered use cases.

For more information about H2OVL Mississippi 2B and 0.8B and other H2O models, visit www.h2o.ai.

About H2O.ai

Founded in 2012, H2O.ai is at the forefront of the AI movement to democratize Generative AI. H2O.ai’s open-source Generative AI and Enterprise h2oGPTe, combined with Document AI and the award-winning autoML Driverless AI, have transformed more than 20,000 global organizations and over half of the Fortune 500, including AT&T, Commonwealth Bank of Australia, Chipotle, ADP, Workday, Progressive Insurance, and AES. H2O.ai’s AI for Good program supports nonprofit groups, foundations, and communities in advancing education, healthcare, and environmental conservation. With a vibrant community of 2 million data scientists worldwide, H2O.ai is on a mission to co-create AI applications that are valuable to all users.

H2O.ai has fundraised $256 million from investors including Commonwealth Bank, Nvidia, Goldman Sachs, Wells Fargo, Capital One, Nexus Ventures, and New York Life.

Contacts

Media Contact
Betty Candel
VP Marketing
betty.candel@h2o.ai

Contacts

Media Contact
Betty Candel
VP Marketing
betty.candel@h2o.ai