FocsleFo’c’sle
GD

RT-2-Mobile

11%
by Google DeepMind

Mobile-class RT-2 distillation under DeepMind's research license.

Vision-Language-ActionCustomMIXEDvlarobotics
19K downloads 320 deploymentsUpdated Feb 14, 2028
Headline:31.8ms · NVIDIA Jetson Thor · MIXED
OverviewBenchmarks3Sim ResultsDeploy3FilesDiscussion23

About this model

Mobile-class RT-2 distillation under DeepMind's research license.

Authored by google-deepmind. Curated into the Fo’c’sle reference set on 2028-02-14. All cross-chip benchmarks below were collected in matched-pair runs in the HIL lab using the same input pipeline, same upstream preprocessing, and the same downstream consumer. See the methodology page for the full protocol.

Task
Vision-Language-Action
Parameters
5.5 B
Benchmarked on
3 chips
Deployments
320

Architecture

Vision-Language-Action policy
Inferred from upstream weights · simplified
RGB camsProprioGoal textVLM backboneAction headDiscretizerJoint cmds

Headline benchmarks

NNVIDIA Jetson ThorMIXED
39.5ms p50
455 tok/s98.8% acc113.6 W
NNVIDIA Jetson AGX OrinMIXED
83.0ms p50
217 tok/s98.6% acc48.2 W
QQualcomm QCS8550MIXED
322.9ms p50
56 tok/s99.3% acc11.7 W

Training data

Pretrained on the upstream maintainer’s released checkpoint. Edge-distillation pass uses 2.4M frames from the Fo’c’sle distillation corpus (consented public data + opt-in publisher contributions). Quantization-aware fine-tune uses 320K calibration samples drawn from the target task’s eval domain.

  • Pretraining corpus: upstream maintainer release
  • Distillation corpus: 2,400,000 frames
  • Calibration set: 320,000 samples (per task)
  • Eval set: standard benchmark + matched-pair HIL runs