Robot Manipulation on Google Robot variant evaluation SimplerEnv benchmark leaderboard · SOTA2 Research

Robot Manipulation on Google Robot variant evaluation SimplerEnv

98Pick Coke Success RateAsyncVLA

Evaluation Results

Method	Links
AsyncVLA2025.11		98	82.3	70.5	50.4	75.3
π02025.11		97.9	78.7	62.3	46.6	71.4
ThinkAct2025.11		92	72.4	50	—	—
AsyncVLA2025.11		91.6	81.7	58	26	64.3
π02025.11		90.1	80.7	27.6	20.5	54.7
SpatialVLA2025.11		88	72.7	41.8	6.3	52.2
SpatialVLA2025.11		86	77.9	57.4	0	55.3
Discrete-Diffusion VLA2025.11		85.4	67.5	60.6	—	—
ThinkAct2025.11		84	63.8	47.6	—	—
Discrete-Diffusion VLA2025.11		82.5	64.6	23.6	—	—
RT-2-X2025.11		82.3	79.2	35.3	20.6	54.4
RT-2-X2025.11		78.7	77.9	25	7.4	47.3
MolmoAct2025.11		77.7	77.1	60	—	—
π0-FAST2025.11		77.6	68.2	31.3	0	44.3
RoboVLM2025.11		77.3	61.7	43.5	24.1	51.7
MolmoAct2025.11		76.1	61.3	78.8	—	—
RoboVLM2025.11		75.6	60	10.6	0	36.6
π0-FAST2025.11		75.3	67.5	42.9	0	46.4
Magma2025.11		75	53	58.9	8.3	48.8
Magma2025.11		68.6	78.5	59	24	57.5
TraceVLA2025.11		60	56.4	31	0	36.9
TraceVLA2025.11		28	53.7	57	0	34.7
Octo-Base2025.11		17	4.2	22.7	0	11
Octo-Base2025.11		0.6	3.1	1.1	0	1.2