Does π0.5 run on Franka Research 3?
Capabilities
| Capability | State | Notes | Evidence |
|---|---|---|---|
| Train / fine-tune | works | ev-hub-bekhzod-pi05, ev-card-msbutt1-pi05-fr3 | |
| Real-robot inference | works with caveats | Guarded live evaluation on a Franka Research 3 via OpenPI serve_policy; no success rate published. | ev-card-msbutt1-pi05-fr3, ev-repo-msbutt1-fr3 |
| Simulation inference | no verified evidence | — | |
| Teleop + record | no verified evidence | — |
Caveats
- Community-reported only; not documented upstream or maintainer-tested.
- The physical deployment is through OpenPI (pi05_droid config, 15 Hz: seven FR3 joint velocities plus gripper closedness) with the authors' guarded client; the LeRobot evidence here is a training checkpoint only.
- No aggregate success metrics are published; the authors withhold them until their v3_12000 validation ledger is reviewed, and the only public rollout evidence is a linked demo video.
- The checkpoint is tied to one cube pick-and-place workcell (two RealSense cameras, its calibration and safety guards) and the authors say not to use it on another workcell without commissioning.
Evidence 3 sources; the cell shows the highest tier
- T3 · community reportedmodel-cardsource 2026-07-22verified 2026-10-25 (today)
OpenPI checkpoint card: pi0.5-DROID fine-tuned on a Franka Research 3 (FR3, not Panda) cube pick-and-place dataset; deployment uses a local FR3 reference streamer, workspace/joint guards, grasp verification and an E-stop; aggregate success metrics intentionally not reported.
“This checkpoint is a pi0.5-DROID policy fine-tuned for real-world pick-and-place with a Franka Research 3 (FR3) robot.”
https://huggingface.co/msbutt1/pi0.5-fr3-v3-pick-place/blob/46173e13f41e5dc8dba8e1ee29f4fbf71a95c236/README.md - T3 · community reportedsource-codesource 2026-10-02verified 2026-10-25 (today)
The authors' FR3 repo documents serving this checkpoint with OpenPI and running it from a libfranka/FCI robot workstation, and links a Live Policy Evaluations wiki page whose 'Pick and Place' entry is a demo video (title read via oEmbed; footage not independently reviewed). It also warns the bracket/bolt collection config uses a different absolute-joint action schema.
“The released cube policy and the dated bracket evaluation use pi0.5-DROID joint-velocity actions: seven FR3 joint velocities plus gripper closedness at 15 Hz.”
https://github.com/msbutt1/franka-fr3-robot-learning/blob/14760a4f6738e393ea9815d539b8ac05e4ad88a2/README.md - T3 · community reportedhub-checkpointsource 2026-07-17verified 2026-10-24 (1d ago)
config.json type=pi05 (pretrained_path lerobot/pi05_base); train_config dataset Bekhzod/fr3_pickplace_fixed_goal_robustness_recovery_merged_data (350 episodes, front + top cameras), whose meta/info.json robot_type=FR3WMS. Auto-generated card, no rollout results.
https://huggingface.co/Bekhzod/Pi05_fr3_pickplace_fixed_goal_robustness_recovery_data_model
π0.5 inference speed by hardware robot-independent · inference cost only
| Compute | Runtime | Precision | Cams | p50 / p95 ms | Inference | Platform | Source |
|---|---|---|---|---|---|---|---|
| RTX 4090 | llama.cpp | fp16 | 2 @224x224 | 51.7 / — | 19.34 Hz | —conditionsboard: NVIDIA RTX 4090 · ggml_cuda_f16_f32_out: 0 (README preset for RTX 4090) Inference cost only; not task success. Full VLM + action-expert pass (ViT + LLM + 10-step action expert), LIBERO inputs (agent-view + wrist at 224x224). From the EagleVLA-Edge README RTX 4090 table (P50), PR #6 configuration. The README's naive π0.5 baseline on this GPU: 162.9 ms (not filed as a separate run: the baseline precision is not stated). | T3 · community reported |
| RTX 4090 | llama.cpp | fp16 | 2 @224x224 | 58.5 / — | 17.09 Hz | —conditionsboard: NVIDIA RTX 4090 · ggml_cuda_f16_f32_out: 0 (README preset for RTX 4090) Inference cost only; not task success. Full VLM + action-expert pass (ViT + LLM + 10-step action expert), LIBERO inputs (agent-view + wrist at 224x224). From the EagleVLA-Edge README RTX 4090 table (P50). The README's naive π0.5 baseline on this GPU: 162.9 ms (not filed as a separate run: the baseline precision is not stated). | T3 · community reported |
| Jetson AGX Thor 128GB | llama.cpp | fp16 | 2 @224x224 | 132.8 / — | 7.53 Hz | —conditionsboard: Jetson Thor (paper/README: "Jetson Thor") Inference cost only; not task success. Full VLM + action-expert pass (ViT + LLM + 10-step action expert), LIBERO inputs (agent-view + wrist at 224x224). Later configuration from the EagleVLA-Edge README (PR #6); the README notes an accuracy caveat for the two-switch GEMM preset on Thor. Naive PyTorch π0.5 on the same board per the paper: 457.9 ms (2.18 Hz) (not filed as a separate run: the baseline precision is not stated). | T3 · community reported |
| Jetson AGX Thor 128GB | llama.cpp | fp16 | 2 @224x224 | 202.8 / — | 4.93 Hz | —conditionsboard: Jetson Thor (paper/README: "Jetson Thor") Inference cost only; not task success. Full VLM + action-expert pass (ViT + LLM + 10-step action expert), LIBERO inputs (agent-view + wrist at 224x224). Paper's 'control frequency' (10.25 Hz) is 1/reaction time with VLM-skipping scheduling; not executed actions/s. Board inferred to be the 128 GB T5000 class from the paper's 130 W max-power entry for Thor (T4000 is 40–70 W). Naive PyTorch π0.5 on the same board per the paper: 457.9 ms (2.18 Hz) (not filed as a separate run: the baseline precision is not stated). | T3 · community reported |
| Jetson AGX Orin 64GB | llama.cpp | fp16 | 2 @224x224 | 271.1 / — | 3.69 Hz | power_mode 50W (paper label "Orin-50W")conditionsboard: Jetson Orin, 'Orin-50W' power mode (AGX Orin 64GB inferred, see ev-orin-board-inferred) Inference cost only; not task success. Full VLM + action-expert pass (ViT + LLM + 10-step action expert), LIBERO inputs (agent-view + wrist at 224x224). Later configuration from the EagleVLA-Edge README (time cache + cuBLASLt GEMM tuning, PR #6); not in the paper. Naive PyTorch π0.5 on the same board per the paper: 1420.8 ms (0.70 Hz) (not filed as a separate run: the baseline precision is not stated). | T3 · community reported |
| Jetson AGX Orin 64GB | llama.cpp | fp16 | 2 @224x224 | 374.5 / — | 2.67 Hz | power_mode 50W (paper label "Orin-50W")conditionsboard: Jetson Orin, 'Orin-50W' power mode (AGX Orin 64GB inferred, see ev-orin-board-inferred) Inference cost only; not task success. Full VLM + action-expert pass (ViT + LLM + 10-step action expert), LIBERO inputs (agent-view + wrist at 224x224). Paper's 'control frequency' (6.93 Hz) is 1/reaction time with confidence-based scheduling that skips the VLM; it is not executed actions/s and is not recorded as control_rate_hz. Naive PyTorch π0.5 on the same board per the paper: 1420.8 ms (0.70 Hz) (not filed as a separate run: the baseline precision is not stated). | T3 · community reported |
| Jetson Orin Nano 8GB (Super) | TensorRT | fp16 | 2 @640x480 | 438.47 / 438.93 | 2.28 Hz | jetpack 7.2.1 · power_mode MAXN_SUPER, clocks pinned with jetson_clocks (GPU 1020 MHz, CPU 1728 MHz) · tensorrt 10.16.2.10conditionsboard: Jetson Orin Nano Engineering Reference Developer Kit Super (8 GB) · l4t: R39.2.1 · os: Ubuntu 24.04.4 LTS · python: 3.12.3 · desktop: running Inference cost only; not task success. 300 s run on synthetic 640x480 observations; 10 denoising steps; latency is runtime-only over a pre-materialized observation ring. Whole-board RAM in use 6.39 GB (unified memory, desktop running). Actions were gated against the stock policy (cosine >= 0.99999) before timing. | T3 · community reported |
| Jetson Orin Nano 8GB (Super) | TensorRT | fp16 | 3 @640x480 | 643.39 / 644.92 | 1.55 Hz | jetpack 7.2.1 · power_mode MAXN_SUPER, clocks pinned with jetson_clocks (GPU 1020 MHz, CPU 1728 MHz) · tensorrt 10.16.2.10conditionsboard: Jetson Orin Nano Engineering Reference Developer Kit Super (8 GB) · l4t: R39.2.1 · os: Ubuntu 24.04.4 LTS · python: 3.12.3 · desktop: running Inference cost only; not task success. 300 s run on synthetic 640x480 observations; 10 denoising steps; latency is runtime-only over a pre-materialized observation ring. Whole-board RAM in use 6.18 GB (unified memory, desktop running). Actions were gated against the stock policy (cosine >= 0.99999) before timing. | T3 · community reported |
Embed this cell
Badge for READMEs and model cards. It always shows the current state.
[](https://caniembody.pages.dev/compat/franka-fr3/pi05)<iframe src="https://caniembody.pages.dev/embed/franka-fr3/pi05" width="380" height="132" style="border:0" title="π0.5 on Franka Research 3 | CanIEmbody"></iframe>