Performance#
RPent Leaderboard
LIBERO-PRO
| Method | Overall | Spatial Task | Spatial Swap | Object Task | Object Swap | Goal Task | Goal Swap | Long Task | Long Swap |
|---|---|---|---|---|---|---|---|---|---|
| RPent / GPT-6 Astra[4]Codex · low · reasoning | 92.63% (741/800) | 100.00% (100/100) | 98.00% (98/100) | 100.00% (100/100) | 99.00% (99/100) | 88.00% (88/100) | 99.00% (99/100) | 85.00% (85/100) | 72.00% (72/100) |
| RPent / DeepSeek V4.1 Flash[5]Codex · xhigh · reasoning | 90.00% (720/800) | 98.00% (98/100) | 98.00% (98/100) | 100.00% (100/100) | 100.00% (100/100) | 90.00% (90/100) | 93.00% (93/100) | 74.00% (74/100) | 67.00% (67/100) |
| RPent / DeepSeek V4.1 Flash[6]Codex · xhigh · reasoning | 88.63% (709/800) | 99.00% (99/100) | 90.00% (90/100) | 99.00% (99/100) | 100.00% (100/100) | 79.00% (79/100) | 98.00% (98/100) | 70.00% (70/100) | 74.00% (74/100) |
| RPent Flash Mode[2]Molmo2-8B · visual localization | 72.63% (581/800) | 79.00% (79/100) | 70.00% (70/100) | 86.00% (86/100) | 93.00% (93/100) | 74.00% (74/100) | 65.00% (65/100) | 60.00% (60/100) | 54.00% (54/100) |
| RPent / Qwen3.6 27Bno-reasoning | 70.63% (565/800) | 82.00% (82/100) | 78.00% (78/100) | 83.00% (83/100) | 84.00% (84/100) | 68.00% (68/100) | 68.00% (68/100) | 61.00% (61/100) | 41.00% (41/100) |
| RPent / GPT-5.5Codex · xhigh · reasoning | 75.13% (601/800) | 81.00% (81/100) | 69.00% (69/100) | 94.00% (94/100) | 91.00% (91/100) | 75.00% (75/100) | 66.00% (66/100) | 70.00% (70/100) | 55.00% (55/100) |
| RPent / GPT-6 Motor Only[3]Codex · low · reasoning | — | — | — | — | — | — | — | 38.00% (38/100) | — |
| π0 | 0.30% | 0.00% | 0.00% | 0.00% | 2.00% | 0.00% | 0.00% | 0.00% | 0.00% |
| π_RLinf | 50.00% | 42.00% (42/100) | 59.00% (59/100) | 71.00% (71/100) | 78.00% (78/100) | 45.00% (45/100) | 42.00% (42/100) | 49.00% (49/100) | 14.00% (14/100) |
| AtomVLA | 6.30% | 1.00% | 16.00% | 0.00% | 10.00% | 11.00% | 2.00% | 9.00% | 1.00% |
| π0.5 | 11.00% | 1.00% | 20.00% | 1.00% | 17.00% | 2.00% | 38.00% | 1.00% | 8.00% |
| MolmoAct | 1.50% | 0.00% | 0.00% | 0.00% | 6.00% | 0.00% | 0.00% | 6.00% | 0.00% |
| X-VLA | 3.80% | 0.00% | 0.00% | 8.00% | 2.00% | 9.00% | 1.00% | 10.00% | 0.00% |
| Cap-X | — | 14.00% | 12.00% | 18.00% | 22.00% | 17.00% | 26.00% | — | — |
| RATS | — | 31.00% | 29.00% | 63.00% | 61.00% | 36.00% | 43.00% | — | — |
| RPent / Opus-4.7Claude Code · max · reasoning | 82.40% (659/800) | — | — | — | — | — | — | — | — |
| ASPIRE[1] | 61.36% | 60.00% (300/500) | 51.00% (255/500) | 95.00% (475/500) | 98.00% (490/500) | 45.00% (225/500) | 81.00% (405/500) | 38.30% (192/500) | 22.60% (113/500) |
* [1] ASPIRE: Long Task and Long Swap use zero-shot transfer from the LIBERO-90 skill library.
* [3] RPent / GPT-6 Motor Only uses motor-only control: it directly outputs end-effector pose increments and gripper commands through execute_action, without invoking VLA / primitives or loading memory.
* [4] GPT-6 Astra (memory-enabled configuration): Long Task/Swap and the other six suites use separate memory-file snapshots frozen after their respective exploration phases, with no updates during evaluation. Overall combines two non-overlapping batches: Long 157/200 plus the other suites 584/600, giving 741/800 (92.63%); the 800 episodes do not share a single memory snapshot.
* [5] DeepSeek V4.1 Flash evaluated with memory explored by GPT-6 Astra.
* [6] DeepSeek V4.1 Flash evaluated with its own seed-0 exploration memory.
LIBERO
| Method | Overall | Spatial | Object | Goal | Long |
|---|---|---|---|---|---|
| RPent / Opus-4.7Claude Code · max · reasoning | 96.00% (384/400) | 97.00% (97/100) | 100.00% (100/100) | 94.00% (94/100) | 93.00% (93/100) |
| OpenVLA | 76.50% | 84.70% | 88.40% | 79.20% | 53.70% |
| NORA | 79.50% | 85.60% | 89.40% | 80.00% | 63.00% |
| π0 | 94.20% | 96.80% | 98.80% | 95.80% | 85.20% |
| π_RLinf | 95.30% | 99.00% (99/100) | 96.00% (96/100) | 97.00% (97/100) | 89.00% (89/100) |
| AtomVLA | 97.00% | 96.40% | 99.60% | 97.60% | 94.40% |
RoboCasa365 · Target50
| Method | Overall | Atomic-Seen | Composite-Seen | Composite-Unseen |
|---|---|---|---|---|
| RPent / GPT-6 AstraCodex · high · reasoning | 59.20% (201/340) | 87.78% (158/180) | 43.75% (35/80) | 42.50% (34/80) |
| RPent / GPT-5.5Codex · xhigh · reasoning | 57.10% (194/340) | 92.00% (166/180) | 61.00% (49/80) | 13.80% (11/80) |
| RPent / Opus-4.7Claude Code · max · reasoning | 48.60% (165/340) | 79.40% (143/180) | 47.50% (38/80) | 15.00% (12/80) |
| π0 | 14.80% | 34.60% | 6.10% | 1.10% |
| π0.5 | 16.90% | 39.60% | 7.10% | 1.20% |
| RLDX-1 | 30.00% | 60.00% | 21.30% | 5.00% |
| WorldDreamer | 35.30% | 66.30% | 26.70% | 9.00% |
| Xiaomi-Robotics-1 | 57.40% | 80.20% | 57.10% | 32.10% |
RoboTwin
| Method | C2R |
|---|---|
| RPent / GPT-5.5Codex · xhigh · reasoning | 62.40% (156/250) |
| RPent / Opus-4.7Claude Code · max · reasoning | 58.40% (146/250) |
| π0.5 | 47.90% |
| GR00T-N1.7 | 20.70% |
| StarVLA | 10.60% |
| LingBot-VLA | 50.40% |