Plurai has secured $13 million in seed funding led by Team8 to develop a simulation layer that improves the reliability of AI agents. The startup uses synthetic data to train these agents, creating a digital twin of a company’s systems for open-weight models. Co-founders Ilan Kadar and Elad Levi bring experience from Nexar and Waymo to the project.
Reah Miyara, a researcher at Google DeepMind, is an angel investor in the round. The platform aims to reduce the costs of running advanced models while maintaining high performance standards. Tests conducted with Nvidia and Nebius demonstrated that Plurai’s models outperform midtier models at a lower cost.
LiveBench 2.0 results show that Claude Fable 5.2 and ChatGPT-6 Astra successfully completed 82.9 and 75.6 tasks respectively. This funding will accelerate the deployment of digital twins for enterprise systems. Companies looking to optimise their AI infrastructure should watch Plurai’s progress closely.
