Franka FR3 / Panda

Facts

Form: Single arm 0 Source
DoF: 7 1 Source
End effector: Franka Hand parallel gripper 1 Source
SDK / ROS: frankarobotics.github.io ↗ 1 Source
Public price (CNY): No price listed on official site 0 Source
Data formats: libfranka / FCI（1 kHz） · ROS 2 · DROID（RLDS） 1 Source

Sources · 1 sources

dof,end_effector,sdk_url,data_formats https://franka.de/products/franka-research-3 Official page · Fetched 2026-09-05 · No snapshot stored yet Official site: 7 Degrees of freedom; Payload 3 kg; 855 mm; libfranka open source C++; FCI 1 kHz; ROS 2

Models declaring this form as a target

Vendor claims, not verified compatibility. Verified states appear in the matrix below.

LingBot-VLA 2.0 Robbyant (Ant Group) Vendor-declared target form

π0 (pi-zero) Physical Intelligence Vendor-declared target form

π0.5 Physical Intelligence Vendor-declared target form

SmolVLA Hugging Face (LeRobot) Vendor-declared target form

OpenVLA Stanford / UC Berkeley / TRI et al. (OpenVLA team) Vendor-declared target form

X-VLA Tsinghua University AIR (2toINF) Vendor-declared target form

Octo UC Berkeley RAIL et al. Vendor-declared target form

RDT-1B Tsinghua University TSAIL (thu-ml) Vendor-declared target form

CogACT Microsoft Research Vendor-declared target form

SpatialVLA Shanghai AI Laboratory et al. Vendor-declared target form

UniVLA OpenDriveLab (HKU et al.) Vendor-declared target form

MolmoAct Allen Institute for AI（Ai2） Vendor-declared target form

RynnVLA-001 Alibaba DAMO Academy Vendor-declared target form

InternVLA-M1 Shanghai AI Laboratory InternRobotics Vendor-declared target form

GraspVLA Peking University EPIC Lab / Galbot Vendor-declared target form

EO-1 IPEC-COMMUNITY (Shanghai AI Laboratory affiliated) Vendor-declared target form

NORA Declare Lab (SUTD) Vendor-declared target form

VLA-Adapter OpenHelix team Vendor-declared target form

HPT（Heterogeneous Pre-trained Transformers） MIT (Lirui Wang, Kaiming He et al.) Vendor-declared target form

LAPA KAIST / University of Washington et al. (Latent Action Pretraining) Vendor-declared target form

LLaVA-VLA OpenHelix team Vendor-declared target form

RoboVLMs（KosMos P.H.） Tsinghua University / ByteDance Research / CASIA et al. Vendor-declared target form

Magma-8B Microsoft Research Vendor-declared target form

V-JEPA 2-AC Meta FAIR Vendor-declared target form

GR-1 ByteDance Research Vendor-declared target form

Spirit v1.5 Spirit AI Vendor-declared target form

Galaxea G0.5 Galaxea Vendor-declared target form

InternVLA-A1.5 Shanghai AI Laboratory InternRobotics Vendor-declared target form

VLA-0 NVIDIA Research Vendor-declared target form

MiniVLA Stanford ILIAD Vendor-declared target form

GR00T N1.5 3B NVIDIA Vendor-declared target form

Compatibility on this embodiment

	Franka FR3 / Panda
Isaac GR00T N1.7
GR00T N1.6 3B
LingBot-VLA 2.0
GO-1 (Genie Operator-1)
π0 (pi-zero)
π0.5
SmolVLA
OpenVLA
X-VLA
Octo
RDT-1B
CogACT
SpatialVLA
UniVLA
MolmoAct
RynnVLA-001
WALL-OSS
InternVLA-M1
GraspVLA
Being-H0
EO-1
LingBot-VLA 4B
NORA
VLA-Adapter
HPT（Heterogeneous Pre-trained Transformers）
LAPA
LLaVA-VLA
RoboVLMs（KosMos P.H.）
Magma-8B
V-JEPA 2-AC
DexVLA
GR-1
Spirit v1.5
Galaxea G0.5
RDT2-VQ
InternVLA-A1.5
VLA-0
MiniVLA
Genie Envisioner（GE-Act）
GR00T N1.5 3B

Legend Officially supportedCommunity reproducedFine-tunable in theoryUnsupportedUnknown Vendor-declared target form 4 / 40 cells with a definite status

Public datasets covering this body

Matched by the robot names or form the dataset states. Scale as stated; blanks are unknowns.

Dataset	Hours	Episodes	Format	Licence
InternData-A1 InternRobotics	7,433	630,000	LeRobot v2.1 and LeRobot v3 (Parquet + MP4)	✗ CC BY-NC-SA 4.0 (HF card); Mod
DROID (Distributed Robot Interaction Dataset) DROID collaboration (13 institutions)	350	76,000	RLDS/TFDS	?
ARIO (All Robots In One) ARIO team (hosted on OpenI, Peng Cheng Laboratory)	—	3,033,188	custom (ARIO unified format)	? CC-BY 4.0 or MIT for ARIO-coll
RoboVerse (roboverse_data) RoboVerseOrg	—	500,000	—	? apache-2.0 (HF license tag); p
PhysicalAI-Robotics-GR00T-X-Embodiment-Sim NVIDIA	—	345,102	LeRobot v2 layout (GR00T LeRobot-compatible: data/chunk-/episode_.parquet + videos/*.mp4 + meta/info.json, modality.json, tasks.jsonl)	✓ cc-by-4.0
RT-1 dataset (Google robot) Google	—	130,000	—	?
RoboMIND X-Humanoid (x-humanoid-robomind)	—	107,000	—	✓ apache-2.0
BridgeData V2 UC Berkeley RAIL	—	60,096	raw (JPEG images, zip downloads)	✓ Creative Commons Attribution 4
MimicGen Datasets NVIDIA (NVlabs)	—	48,000	HDF5 (robomimic)	✓ CC-BY 4.0
CALVIN mees/calvin (Oier Mees et al.)	—	—	—	✓ MIT
ManiSkill3 Demonstrations haosulab (ManiSkill team)	—	—	HDF5 + JSON metadata	? Apache-2.0 (GitHub repository
RLBench Imperial College London (Stephen James et al.)	—	—	custom (CoppeliaSim/V-REP .ttm task files + Python task files; demonstrations generated on demand via motion planners)	✗ Custom Imperial College London
Unitree Robotics open datasets (Hugging Face org) Unitree Robotics	—	—	LeRobot v3 (codebase_version 'v3.0' in sampled dataset G1_Dex1_Wipe_Board)	✓ Apache-2.0 (94 of 160 datasets
UMI (Universal Manipulation Interface) datasets Stanford REAL Lab (real-stanford) / Columbia	—	—	zarr (.zarr.zip)	? MIT (code repository); dataset

Clear direction, grounded judgement.