Skip to content
RobothecaBeta
Log in

SmolVLA

  • AI model
  • Humanoids and robot learning
  • Runs on hardware you can build

A compact vision-language-action model from the LeRobot team that turns camera images and a task in words into arm actions; it trains on one GPU and runs with LeRobot.

Made by Hugging Face.

Source on Other: huggingface.co/lerobot/smolvla_base (opens in a new tab).

Licenses

Code
Apache-2.0 (opens in a new tab)

A person at Robotheca set it on 21 Oct 2026. Checked by hand on 21 Oct 2026.

Commercial use
Yes

What it does

Robotheca has not written what it does yet. Until then, see the project's own page: huggingface.co/lerobot/smolvla_base (opens in a new tab).

Getting started

Difficulty
Intermediate

Maintenance

Status
Active: a commit in the last 90 days
Newest commit
6 Oct 2026

Checked by hand on 21 Oct 2026.

Links

Robots that can run it

  • Robot arm

    SO-101

    A low-cost 3D-printed robot arm built as a leader and follower pair, for teleoperation and for teaching tasks by demonstration with LeRobot.

    • Works with AI
    Level
    Beginner
    Buildability
    98/100
    Ready to build
    Cost
    €226.30

    The model card runs SmolVLA through LeRobot on an SO-100 arm (lerobot-record --robot.type=so100_follower); the SO-101 is the SO-100's successor, set up with LeRobot the same way.

Robotheca lists AI models as a reference: nothing is hosted or run here, and every fact links to where it was read.