Model comparison

act_so101_keyboard_gray_cube_clean11 vs o3

Capability radar, value bars, and side-by-side posture.

← Back to Compare

act_so101_keyboard_gray_cube_clean11
Not scored
value score
o3
93
value score
act_so101_keyboard_gray_cube_clean11 ctx
o3 ctx
200K

Capability radar

Value · context · multimodal · openness · speed posture

Value head-to-head

  • o393

Attribute tape

Fieldact_so101_keyboard_gray_cube_clean11o3
DeveloperTimmyOpenAI
Context200K
ModalitiesText, Code
OpennessProprietary Model
Speed
Price$10/1M in · $40/1M out
Value score93
SummaryA task-specific ACT (Action Chunking with Transformers) imitation-learning policy checkpoint for the SO-101 robotic arm, trained via the Hugging Face LeRobot stack to perform a tabletop keyboard/gray-cube manipulation task. It outputs chunked action sequences from sensory input to control the arm through the learned cleaning behavior.OpenAI reasoning model focused on hard problems.
Open act_so101_keyboard_gray_cube_clean11Open o3