Mantic Secures $25M to Scale AI Forecasting

The big picture: British AI startup Mantic has secured $25 million in seed funding. The investment follows a significant achievement where its forecasting system outperformed all human competitors in the summer 2026 Metaculus Cup, providing a strong validation for its advanced predictive AI.

Why it matters:

• Superhuman Performance: Mantic’s AI system demonstrated superior judgmental forecasting by beating human experts in a major public competition, marking a shift in AI’s capability beyond generative tasks.
• Strategic Investment: The round, led by Radical Ventures and including Microsoft’s M12, signals growing investor confidence in AI systems designed for complex, real-world prediction rather than conventional data modeling.
• New AI Frontier: This funding highlights an emerging area of AI focused on messy, uncertain events in business, geopolitics, and culture, where accurate foresight can translate directly into strategic and financial advantages.

How it works:

• Judgmental Forecasting: Mantic specializes in predicting outcomes for complex, real-world questions that require research, context, and judgment, often weeks or months into the future, rather than relying solely on clean datasets.
• Model Adaptation: The startup leverages and adapts existing foundation models from other AI labs, tailoring them specifically for sophisticated forecasting tasks.
• Continuous Refinement: Mantic’s system is rigorously tested against historical events, scored on its predictions, and continuously refined based on these performance results to enhance accuracy.

The catch: While the Metaculus Cup win is a compelling proof point, a single tournament cannot definitively establish an AI system’s universal reliability across all types of future events. Forecasting performance can vary significantly based on the question’s nature, time horizon, and available information, suggesting Mantic will need to build sustained trust and demonstrate consistent accuracy across diverse, real-world applications to overcome inherent market skepticism in predictive technologies.