Hugging FaceUnited States2025

SmolVLA

An AI model: a vision-language-action model (VLA) takes camera images and an instruction and turns them into movements for a robot. This page describes the software, not a robot.

Robotsnot stated

Key figures as stated by the developer

6 of 6 stated

Model type ⁠A vision-language-action
Controls ⁠B arms and hands
Inputs ⁠A camera language robot state
Access ⁠C open download
Parameters ⁠C 450,000,000
Base model ⁠A SmolVLM2-500M-Video-Instruct

How to read the figures

14kgA30.9lbA: The letter in the grey square after a figure points to the source the figure came from. Robots with several sources carry several letters.
74.5lb: Imperial figures are our own conversion unless the manufacturer states an imperial figure itself. The source letter always belongs to the metric figure — a conversion has no source of its own.

The four states

Versions and notes in the key figures (6)

Model type

Developer wording: "SmolVLA is a compact, efficient Vision-Language-Action (VLA) model designed for affordable robotics".

Controls

Model card, outputs: "continuous actions". The tested robots are arms: the paper collected "three datasets using the SO-100 robot arm".

Inputs

Model card, inputs: "images (multi-view), proprio/state, optional language instruction".

Access

Blog: "SmolVLA-450M is our open-source, compact yet capable VLA model." "Released with full training and inference recipes".

Parameters

Blog: "SmolVLA-450M is our open-source, compact yet capable VLA model."

Base model

Model card: "base_model: HuggingFaceTB/SmolVLM2-500M-Video-Instruct".

Full specifications

9 of 12 fields stated

The letter points to the source at the foot of the page. Fields without a figure are dashed: the developer does not state them — that is not the same as zero, and not the same as no.
Field	Value	Source and caveats
Input and output
Model type	`vision-language-action`	A Developer wording: "SmolVLA is a compact, efficient Vision-Language-Action (VLA) model designed for affordable robotics".
Inputs	`camera` `language` `robot state`	A Model card, inputs: "images (multi-view), proprio/state, optional language instruction".
Controls	`arms and hands`	B Model card, outputs: "continuous actions". The tested robots are arms: the paper collected "three datasets using the SO-100 robot arm".
Degrees of freedom controlled	not stated	Note The only count is for the test arms, not the model: "Each arm has six degrees of freedom". ⁠B
Size and speed
Parameters	450,000,000	C Blog: "SmolVLA-450M is our open-source, compact yet capable VLA model."
Control rate	not stated
Where it runs and what it is built on
Runs where	not stated	Note The developer says "Small enough to run on CPU, train on a single consumer GPU, or even a MacBook!" and, for its asynchronous setup, that "the policy can be run on a remote server, possibly with GPUs"; it does not state where the model runs on a robot. ⁠C
Base model	SmolVLM2-500M-Video-Instruct	A Model card: "base_model: HuggingFaceTB/SmolVLM2-500M-Video-Instruct".
Openness
Access	open download	C Blog: "SmolVLA-450M is our open-source, compact yet capable VLA model." "Released with full training and inference recipes".
License	Apache-2.0	A Model card: "license: apache-2.0".
The developer’s own claims
Learns new tasks	needs retraining	A Developer claim. Model card, intended use: "Base model to fine tune on your specific use case".
Works across robots	yes	C Developer claim: "Can be tested and deployed on very affordable hardware (SO-100, SO-101, LeKiwi, etc.)". Yes only if the developer itself says the model runs on different robot models, for example robots from two manufacturers or two of one manufacturer’s models. Several units of the same robot do not count.

Notes

Hugging Face describes SmolVLA as a compact, efficient Vision-Language-Action model designed for affordable robotics.
The SmolVLA paper appeared on arXiv on 2 June 2025.
The Hugging Face terms of service describe the company as a Delaware corporation.

Sources

A Manufacturer product page. huggingface.co/lerobot/smolvla_base · Retrieved 2026-10-13
B arxiv.org/abs/2506.01844 · Retrieved 2026-10-13
C huggingface.co/blog/smolvla · Retrieved 2026-10-13

One line per source, not per figure: a source has one retrieval date. The letter beside a figure points to the line it came from.