Skip to main content

Model Library

Model Library​

The Model Library lets you easily upload and manage a wide range of AI models.

Register models on top of the BentoML environment and run them as either Serverless or Dedicated as needed, across categories like natural language processing (NLP), computer vision (CV), and speech processing.

You can also version your models, track changes to uploaded models, and roll back to a previous version if needed. This makes model development and operations more reliable and supports continuous improvement.

The Model Library plays a role similar to a model registry: uploaded models can be used inside your organization or shared publicly. This makes it easier to manage research and experimental models systematically and collaborate efficiently.

Browsing Models​

AI model cards

The Model Library screen shows every available AI model as a card. Each card includes:

  • Model name
  • Execution mode (Serverless / Dedicated)
  • Provider (OpenAI / elice / external partners, etc.)
  • Pricing (Input / Output-based billing)
  • Supported capabilities (Text Generation, STT, TTS, etc.)

The following browsing tools are also provided.

Search and filter​

  • Search by model name

  • Filter by execution mode

    • Serverless
    • Dedicated

  • Filter by domain

    • Text / Vision / Speech / Reasoning / Classification, etc.

Use these to quickly find the model you need.

Model Categories​

The Model Library offers AI models in the following categories.

Natural Language Processing​

  • Text Generation: generates new text based on input. Used in chatbots, content generation, and similar applications.
  • Image Text to Text: recognizes text in images and converts or summarizes it. Used for OCR (optical character recognition) and document processing.

Computer Vision​

  • Image Generation: generates new images from input. Used for artistic image generation, data augmentation, and so on.
  • Object Detection: detects specific objects in images or video and identifies their location. Used in autonomous driving, security surveillance, and similar systems.
  • Face Detection: recognizes faces in images or video. Used in face authentication, emotion analysis, and similar applications.

Speech Processing​

  • Text To Speech: converts text into natural-sounding speech. Used in voice assistants, audiobooks, and similar applications.
  • Speech To Text: converts speech into text. Used in automated captioning, voice command systems, and similar applications.

Execution Modes​

Each model is offered in one or both of the following modes.

Serverless​

  • Call instantly without preparing infrastructure
  • Per-call billing

Dedicated​

  • Model is deployed to a dedicated instance
  • Suited to stable performance and sustained traffic

Model Details​

Model detail page

Clicking a model opens its detail page, which shows:

  • Model description and key capabilities
  • Provider and supported domains
  • Available execution modes (Serverless / Dedicated)
  • Pricing
  • Usage examples

Visibility​

  • Organization-only: visible only to members of your organization.
  • Public: visible to all Elice Cloud users.