SmolVLA
Hugging Face LeRobot
Compact open vision-language-action policy designed for practical robot fine-tuning and deployment through the LeRobot ecosystem.
- Code
- Available
- Weights
- Available
- Checkpoint
- Available
- License
- Apache-2.0
Decision summary
Data-efficient fine-tuning
Hardware, latency, dependencies, and checkpoint loading are not pipeline-tested.
Run a checkpoint loading smoke test before marking the model pipeline-tested.
Release history
Source-backed release timing for this canonical model record.
- Hugging Face post
Model series introduced
Release timing is recorded from official model metadata.
Selection readiness
Evidence dimensions for deciding whether this model is ready to inspect, compare, or adopt. Scores describe catalog evidence, not task performance.
Published evaluation
How scores work76/100
Useful
5/6 dimensions scored · 63% confidence
Engineering checks (OBRS) behind this record
Needs Audit
OBRS metadata and reported test evidence. A breakdown behind Selection readiness, not a separate score or an independent OpenBot certification.
- A passing real-hardware test report is required.
- A passing ingestion/pipeline test report is required.
- A passing privacy and provenance review is required.
Why these scores
The strongest decision reasons behind Selection readiness.
Artifact availability
UsefulCheckpoint files were verified in the official repository metadata.
Training and loading reproducibility
UsefulConfiguration files are present; runtime loading is not tested.
Evaluation evidence
UnknownNo structured evaluation evidence has been verified.
Data requirements
Declared loop-data needs, missing evidence, and the next checks that matter.
Observation / ego video
observation
Language intent / task phase
language intent
Action / robot state
actions · robot state
Feedback / correction / failure
task success
Critical gaps
Core categories are represented. Interface alignment and data quality still require verification.
Linked datasets
Signal-level links only. Verify runtime interfaces before use.
EgoWorld
Schema validation for LeRobot v3
provisional fit 75 · confidence 30
2 matched · 2 unknown
EgoStation GoPro Pick-and-Place
Pick-and-place benchmark fixtures
provisional fit 88 · confidence 45
3 matched · 1 unknown
Egocentric Adjust Bottle
LeRobot conversion tests
provisional fit 88 · confidence 45
3 matched · 1 unknown
More evidenceShowHide
Additional data signals
Sim-real / embodiment metadata
robot state
Additional decision dimensions
Access and license are declared by the source.
75 · c85Required signal categories are structured; exact tensor and action interfaces still need verification.
76 · c70Hardware, latency, dependencies, and checkpoint loading are not pipeline-tested.
70 · c55Artifact facts
- release_year
- official_claim
- 2025
- Official SmolVLA announcement, Published June 3, 2025
- model.required_signals
- official_claim
- observation · language intent · actions · robot state · task success
- Official model documentation and curated signal mapping
- repository.code
- metadata_verified
- repository: huggingface/lerobot · default branch: main · archived: false · code file count: 753 · sample files: examples/backward_compatibility/replay.py · examples/dataset/create_progress_videos.py · examples/dataset/load_lerobot_dataset.py · examples/dataset/slurm_compute_rabc.py · examples/dataset/use_dataset_image_transforms.py · examples/dataset/use_dataset_tools.py · examples/isaac_teleop_to_so101/__init__.py · examples/isaac_teleop_to_so101/common.py · examples/isaac_teleop_to_so101/isaac_teleop/__init__.py · examples/isaac_teleop_to_so101/isaac_teleop/base.py · examples/isaac_teleop_to_so101/isaac_teleop/clutch.py · examples/isaac_teleop_to_so101/isaac_teleop/config_isaac_teleop.py · examples/isaac_teleop_to_so101/isaac_teleop/teleop_so101_leader_arm.py · examples/isaac_teleop_to_so101/isaac_teleop/teleop_xr_controller.py · examples/isaac_teleop_to_so101/isaac_teleop/xr_controller_processor.py · examples/isaac_teleop_to_so101/override_reset_pose.py · examples/isaac_teleop_to_so101/record.py · examples/isaac_teleop_to_so101/teleoperate.py · examples/lekiwi/evaluate.py · examples/lekiwi/record.py
- GitHub repository metadata and recursive tree
- model.hub
- official_claim
- declared
- curated model hub link
- model.weight_files
- metadata_verified
- repository: lerobot/smolvla_base · count: 3 · sample files: model.safetensors · policy_postprocessor_step_0_unnormalizer_processor.safetensors · policy_preprocessor_step_5_normalizer_processor.safetensors
- Hugging Face model metadata siblings
- model.checkpoint_files
- metadata_verified
- repository: lerobot/smolvla_base · count: 3 · sample files: model.safetensors · policy_postprocessor_step_0_unnormalizer_processor.safetensors · policy_preprocessor_step_5_normalizer_processor.safetensors
- Hugging Face model metadata checkpoint files
- model.config_files
- metadata_verified
- repository: lerobot/smolvla_base · count: 1 · sample files: config.json
- Hugging Face model metadata siblings
- model.training_data
- official_claim
- source: LeRobot Community Datasets · curated datasets: 487 · frames: approximately 10 million · frame rate: 30 FPS · primary embodiment: SO100
- SmolVLA blog, training dataset
- license.code
- metadata_verified
- Apache-2.0
- GitHub repository SPDX license metadata
- license
- metadata_verified
- Apache-2.0
- GitHub repository SPDX license metadata
- release_timing
- metadata_verified
- 2025-06-03T00:00:00.000Z
- Official SmolVLA announcement, Published June 3, 2025
- discovery.source
- secondary_claim
- name: worldbench/awesome-embodied-data-pyramid · trust: discovery_only · revision: 0568599d0619f20946e8089f6a941d0d9e30b690
- #table-embodied-foundation-models-vla-wam, table 17, row 18
- discovery.raw_fields
- secondary_claim
- Time: 2025.6 · Method: SmolVLA · Institution: Hugging Face · Project: [](https://huggingface.co/lerobot/smolvla_base) · Model: VLA · Data: ![Real][data-real]
- #table-embodied-foundation-models-vla-wam, table 17, row 18
- model.reported_release_time
- secondary_claim
- 2025.6
- #table-embodied-foundation-models-vla-wam, table 17, row 18
- model.type
- secondary_claim
- VLA
- #table-embodied-foundation-models-vla-wam, table 17, row 18
- model.training_data_layers
- secondary_claim
- real_robot
- #table-embodied-foundation-models-vla-wam, table 17, row 18
- model.institution
- secondary_claim
- Hugging Face
- #table-embodied-foundation-models-vla-wam, table 17, row 18
- catalog.curation
- secondary_claim
- tier: editorial_focus · collection: WorldBench Awesome Embodied Data Pyramid · policy: human_curated_priority
- #table-embodied-foundation-models-vla-wam, table 17, row 18
Additional references
OpenBot notes
- Strong fit for OpenBot's model-readiness view because public weights make the training path concrete.
- Catalog should surface whether a dataset can be converted into LeRobot-style episodes.