Data as of Sep 17, 2026 · Based on 3,310,041 AI responses across 10,525 prompts · See how Parse measures this
Octoocto-models.github.ioOcto is an open-source, generalist robot policy that uses a transformer-based diffusion model to control manipulation across diverse robots, sensors, and actions, supporting both natural language instructions and goal image conditioning. It is pretrained on about 800k episodes from the Open X-Embodiment dataset and comes in Octo-Small (27M) and Octo-Base (93M), designed for flexible finetuning to new observations and action spaces. Octo has been evaluated on multiple real-robot setups, showing strong out-of-the-box capabilities and substantial gains when finetuned across a range of tasks and embodiments.
Parse Score
Excerpts where Octo appeared in the AI's answer
Octo : A modular, open-source policy framework for generalist robotic manipulation
Octo : An open-source, generalist Transformer-based robotic policy trained on the large-scale RT-X multi-robot dataset.
