Plurai Raises $13 Million to Sharpen AI Agents With Simulations
Updated
Updated · TechCrunch · Nov 11
Plurai Raises $13 Million to Sharpen AI Agents With Simulations
1 articles · Updated · TechCrunch · Nov 11
Summary
$13 million in seed funding will help Plurai build a simulation layer for AI agents, using synthetic data and digital twins of a company’s systems to improve reliability before deployment.
Current agents still struggle on complex work: LiveBench 2.0 showed Claude Fable 5.2 and ChatGPT-6 Astra scoring 82.9 and 75.6 out of 100 tasks, while EON found Claude Opus 4.8 reached 76.8% accuracy.
Plurai maps possible agent workflows across tools, knowledge bases and databases, then uses those simulations to test edge cases and fix failures such as policy violations, multi-dataset errors and context overload.
The startup says that process can also post-train small open-weight models for specific customers; tests with Nvidia and Nebius showed performance above mid-tier frontier models at much lower cost.
Founded by former autonomous-driving engineers from Nexar, Plurai is betting that the simulation methods used to make self-driving systems safer can also make enterprise AI agents dependable and cheaper.