OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Unilm

Unilm is a public Microsoft research repository containing Python implementations and related projects for large-scale self-supervised pre-training across tasks, languages and modalities. It covers foundation architectures, language and multilingual models, document AI, speech, vision and multimodal systems.

Independent OpenViST project profile, based on linked sources.

Unilm — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Unilm when you need a research-oriented starting point for studying or prototyping foundation-model methods across language, documents, speech, vision or multimodal data. Choose the relevant subproject rather than treating the repository as a single application.

Who it is for

  • AI researchers studying foundation-model architectures and pre-training
  • Engineers prototyping language, document, speech, vision or multimodal research workflows
  • Teams evaluating open-source implementations for self-hosted experimentation

Documented uses

  • Explore self-supervised pre-training methods across multiple AI modalities
  • Prototype multilingual language and machine-translation research
  • Build document-AI experiments for scanned documents and PDFs
  • Investigate speech representation learning, automatic speech recognition and text-to-speech
  • Evaluate multimodal foundation-model architectures
Before you choose

Requirements & limits

The repository is public, active, open source and MIT licensed. It is intended for self-hosted use and is identified for Linux, Windows and macOS. No account or pricing information is provided.

  • The repository covers many separate research areas, so implementation requirements depend on the selected subproject.
  • Python is the primary language and self-hosted experimentation is the documented operating model.
  • Document AI coverage includes text, layout or format and image inputs for scanned documents and PDFs.
  • Language coverage includes multilingual and cross-lingual models, including projects described for more than 100 languages.
  • The collection includes both model implementations and foundation-architecture research, making project-level documentation important before adoption.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
2 October 2026
Original website title
microsoft/unilm · Official homepage title · retained in its source language
How we review sourcesSuggest a correction