European roboticists today has released a powerful open source artificial intelligence model that acts as a brain for industrial robots, helping them grasp and manipulate objects with new dexterity.
The new model, SPEAR-1, was developed by researchers at the Institute of Informatics, Artificial Intelligence and Technology (INSAIT) in Bulgaria. This could help other researchers and startups create and experiment with smarter hardware for factories and warehouses.
Just as open source language models have allowed researchers and companies to experiment with generative AI, Martin Vechev, a computer scientist at INSIAT and ETH Zurich, says SPEAR-1 should help roboticists experiment and iterate quickly. “Open models are crucial to advancing embodied AI,” Vechev told WIRED ahead of the release.
SPEAR-1 differs from existing basic robot 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 on vision language models (VLMs) which have a broad but limited understanding of the physical world, as training tends to come from labeled 2D images. “Our approach addresses the mismatch between the 3D space in which the robot operates and the knowledge of the VLM which constitutes the heart of the basic robotic model,” explains Vechev.
SPEAR-1 performs about as well as commercial baseline models designed to operate robots, when measured on RoboArena, a benchmark that tests a model’s ability to get a robot to do things like squeeze a bottle of ketchup, close a drawer, and staple pieces of paper together.
The race to make robots smarter is already worth billions of dollars. The commercial potential of generally high-performance robots has given rise to well-funded startups, including Skild and Generalist, in addition to physical intelligence. SPEAR-1 is almost as good as Pi-0.5 from Physical Intelligence, a billion-dollar startup founded by a team of leading robotics researchers.
SPEAR-1 suggests that the quest to build smarter robots could involve both closed models like those from OpenAI, Google, and Anthropic, as well as open source variants like Llama, DeepSeek, and Qwen.
However, robotic 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 certain objects from a table. In practice, however, the model will need to be retrained from scratch if a different type of robotic arm is used or if the object or environment is changed.
