Modelmultimodalvv1.5

Octo

by UC Berkeley (Academic) · open-source · Last verified 2026-03-17

Octo is an open-source generalist robot policy model from UC Berkeley that can be fine-tuned to new robot setups with minimal data, trained on 800K robot trajectories from the Open X-Embodiment dataset. It uses a flexible transformer architecture that supports language instructions, image observations, and diverse action spaces across different robot morphologies.

https://octo-models.github.io/ ↗

C+

C+—Average

Adoption: CQuality: AFreshness: B+Citations: BEngagement: F

Specifications

License: Apache 2.0
Pricing: open-source
Capabilities: generalist-robot-policy, few-shot-robot-adaptation, multi-embodiment-control, language-conditioned-manipulation
Integrations: JAX, HuggingFace, ROS
Use Cases: robot manipulation research, few-shot robot adaptation, cross-embodiment transfer, academic robotics labs
API Available: Yes
Parameters: ~93M
Context Window: N/A
Modalities: vision, text, action
Training Cutoff: 2024
Tags: robotics, open-source, transformer, manipulation, generalist
Added: 2026-03-17
Completeness: 100%

Index Score

51.4

Adoption

Quality

Freshness

Citations

Engagement

Explore the full AI ecosystem on Agents as a Service