bytedance / bytedance/UI-TARS

Request for Official MLX Support for UI-TARS

Open
#193 4 comments 5 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
11.5k
Forks
877
PR merge metrics
No merged PRs in 30d

Description

First of all, thank you for this amazing model - it works incredibly well and covers almost all our needs for UI testing.

I’m wondering if you have any plans to officially support the MLX format so it can run natively on Apple hardware?

Community-converted models, such as [UI-TARS-1.5-7B-6bit](https://huggingface.co/mlx-community/UI-TARS-1.5-7B-6bit), already perform quite well - even on mid-range MacBooks. However, there seem to be issues with coordinate generation, which currently fails consistently.

Some major players, like Google with Gemma, have already started releasing official MLX-compatible models. It would be amazing to see official MLX support for UI-TARS as well.

Image

Thanks again for your great work!

Contributor guide

No contributing guide indexed for this repository

Research direction

The issue names no repository files, tests, or entry points. Start by examining the linked MLX-community model and reproducing its coordinate-generation failure on Apple hardware; done means official MLX support runs natively and produces correct coordinates.

Written by the indexing model from the issue text.

Assessment

Tech stack
macos, python
Domain
machine-learning, operating-systems
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.