General Intuition Inc. announced a $220 million funding round on September 29, 2026. The investment values the world model startup at $6.2 billion. Valor Equity Partners, Atreides, Seven Seven Six, Point72, Khosla Ventures, and General Catalyst provided the capital. This funding arrives one year after General Intuition spun out of video sharing startup Medal B.V.
Medal offers a free tool for consumers to capture and record video game footage. General Intuition leverages that recorded footage to train its artificial intelligence models. Developers typically train AI models by feeding them examples of tasks they must perform. Manually recording such specialized footage often demands significant time.
General Intuition builds world models designed to generate synthetic footage for artificial intelligence training projects. The company debuted its newest algorithm, MIRA, in June. The company states that this new AI surpasses earlier models in multiple capabilities.
Read nextEliseAI Raises $350 Million And Reaches $4 Billion ValuationMIRA algorithm renders long workflows
Traditional video generators only create clips lasting a few seconds or minutes. This limitation prevents them from producing synthetic videos of long factory automation workflows. Such constraints reduce their practical value for robot developers.
MIRA runs infinitely without diverging. It can render complex scenes featuring multiple fast-moving objects interacting together. This simulation ability helps with tasks like training robots to avoid collisions.
MIRA operates with high efficiency. General Intuition reports that the model renders 20 frames per second at a resolution of 720 by 576 pixels using a single B200 graphics card.
Latent space processing drives efficiency
The model achieves efficiency through latent diffusion processing. Standard media generators process individual video frames directly. MIRA instead runs calculations on a latent space, which is a compressed data structure representing video frames with lower memory requirements.
Reducing the memory footprint decreases the processing time. MIRA also remains small with only 5.6 billion parameters. This parameter count sits far below frontier models.
MIRA currently functions as a research demonstration rather than a commercial training data generator. In its present state, the model only generates synthetic footage of a single video game. General Intuition nevertheless calls MIRA a stepping stone to physical AI because of its rendering quality and efficiency.
General Intuition is testing a commercial version of the technology alongside a limited group of customers. These customers focus on robotics, simulation, and entertainment use cases. The company launched a waitlist for this commercial offering alongside the funding announcement and plans to hire additional AI researchers with the new capital.
