UI TARS

UI-TARS is an open-source multimodal agent that controls your computer by interpreting screen images and natural-language instructions, helping developers and…

  • AI
  • Development
  • Tools & Utilities
UI-TARS is an open-source multimodal agent that controls your computer by interpreting screen images and natural-language instructions, helping developers and power users automate cursor movement, clicks, and screen-based tasks across applications.

Platform availability

Available on 0 of 10 platforms

  • Unraid (not listed)
  • TrueNAS (not listed)
  • Umbrel (not listed)
  • ZimaOS (not listed)
  • Proxmox (not listed)
  • Helm (not listed)
  • Mac (not listed)
  • Windows (not listed)
  • Linux (not listed)
  • Any server (not listed)

Health score?A 0-100 score computed nightly from four factors: maintained (40%: recent commits and steady activity), popular (25%: GitHub stars, log scale), easy to install (20%: platforms, install notes, setup guides, Docker image) and light to run (15%: minimum RAM, ARM support). When the source code has been checked, a fifth factor, security check (20%: automated scan for committed secrets, unsafe desktop-app settings, known-vulnerable dependencies and project hygiene; any serious finding caps it at 20), is added. Factors without data are left out and the rest are rescaled.

44/100 Fair

Maintained
29last commit 442 days ago
Popular
8111,699 GitHub stars
Easy to install
250 install platforms, install notes, Docker image
Light to run
106 GB minimum RAM
Security check
70no serious findings, 15 minor, CI, tests
GitHub stars
11,699
Open issues
59
Last commit
2025-09-05
License
Apache-2.0
Activity
stale

Checked yesterday - source: GitHub (bytedance/UI-TARS)

Resources & Compatibility

Minimum requirements: 6 GB RAM.

Checked 54 days ago - source: Docker Hub image tags and published docs

First-install notes

  • Desktop version A separate UI-TARS-desktop project is provided for running on local personal devices; use it for desktop deployment rather than the main model repository. source
  • Local inference Local deployment can use vLLM, a model-inference server, for on-device inference, keeping the runtime local and removing dependence on cloud services. source

Fetched 48 days ago - each note links to its source

Alternatives (11)

  • DomoAI - AI-powered video platform offering image-to-video, text-to-image and anime-to-photo conversion in multiple styles including anime, paper ar…
  • Dream Machine - AI-powered system generates high quality, realistic videos from text prompts and images, ensuring accurate physics, character consistency,…
  • Google Flow - AI-powered tool for cinematic video creation featuring text and image prompts, camera controls, asset management, and seamless scene buildi…
  • Google Veo - Generates detailed 1080p HD videos with advanced cinematic controls, language and vision synergy, style matching, robust editing tools, and…
  • Hailuo AI - Automates video production with AI, offering tools like trimming, transitions, and templates to simplify creation across industries.
  • KLING AI - Generates videos and images from text or image prompts with high-resolution options, up to 3-minute length, creative clones, and a collabor…
  • Mochi AI - Provides unmatched realistic motion and fluid human action with prompt adherence for customizable AI-generated video simulations.
  • Runway - Suite offers over 30 AI tools for creating, editing, and modifying images, audio, and video with controls for style, resolution, and conten…
  • Sora - Sora is an OpenAI's text-to-video model (makers of ChatGPT, DALL-E 3 and Whisper).
  • Stable Video Diffusion - Open source generative AI video model transforming text and images into cinematic scenes, offering 14 and 25 frame options with flexible fr…
  • Wan - Open-source video model offering superior performance, generating videos efficiently on consumer GPUs and excelling in text conversion and…

Checked 40 days ago - source: AlternativeTo

Common questions

Do I use the main UI-TARS repository or a separate project for desktop control on a personal device?

Use the separate UI-TARS-desktop project for running on a local personal device. The main model repository is intended for model deployment and inference.

Can UI-TARS be run locally without cloud services?

Local deployment can use vLLM as a model-inference server for on-device inference. This keeps the runtime local and removes dependence on cloud services.

How can UI-TARS Desktop avoid dependence on one model provider?

Self-hosting can integrate Lynkr as a universal LLM gateway. UI-TARS Desktop can then be configured to work with your preferred model backends.

What coordinate detail should I verify when using Qwen 2.5VL-based models?

Qwen 2.5VL-based models use absolute coordinates to ground objects. You should follow the coordinate-processing guidance before using them for screen actions.

Answers sourced from deepwiki.com, gcentral.dev, github.com

Community

Collected 48 days ago - public community threads