European roboticists today has released a powerful open-source artificial intelligence model that works like the brain of industrial robots – helping them grab and manipulate things with fresh dexterity.
fresh model, RULER-1was developed by scientists from the Institute of Informatics, Artificial Intelligence and Technology (INSAIT) in Bulgaria. It could assist other researchers and startups build and experiment with smarter equipment for factories and warehouses.
Just as open-source language models have enabled researchers and companies to experiment with generative AI, Martin Vechev, a computer scientist at INSIAT and ETH Zurich, says SPEAR-1 should assist roboticists experiment and iterate quickly. “Open weight models are critical to the development of embodied artificial intelligence,” Vechev told WIRED ahead of the launch.
SPEAR-1 differs from existing robot foundation models in that it incorporates 3D data into its training mix. This gives the model a better understanding of the physical world, making it easier to understand how objects move in physical space.
Basic robot models are typically built based on vision language models (VLM), which have a broad but circumscribed understanding of the physical world because training is usually based on labeled 2D images. “Our approach solves the problem of discrepancy between the 3D space in which the robot operates and the VLM knowledge that forms the core of the robotic foundation model,” says Vechev.
SPEAR-1 has roughly the same capabilities as commercial entry-level models designed to operate robots on RoboArena, the reference point tests a model’s ability to get a robot to do things like squeeze a ketchup bottle, close a drawer, and staple pieces of paper together.
The race to make robots smarter is already on billions of dollars riding it. The commercial potential of generally capable robots has led to the emergence of well-funded start-ups including Separate AND Generalist Apart from Physical intelligence. SPEAR-1 is almost as good as Pi-0.5 from Physical Intelligence, a billion-dollar startup founded by an all-star team of robotics researchers.
SPEAR-1 suggests that the push to build more clever robots could include both closed models such as OpenAI, Google and Anthropic, and open source variants such as Llama, DeepSeek and Qwen.
However, robot intelligence is still in its infancy. It is possible to train an AI model to operate a robotic arm so that it can reliably pick up specific objects from a table. In practice, however, the model will need to be retrained if a different type of robotic arm is used or if the object or environment is changed.