British AI startup Mantic has raised $25 million in a seed round led by Radical Ventures, with participation from Microsoft's M12 venture fund, Balderton Capital, and Mira Murati's Thinking Machines Lab, to scale a forecasting AI the company says already outperforms top human forecasters on real-world questions. The London-based startup, founded in 2024 by former Google DeepMind researcher Toby Shevlane and engineer Ben Day, plans to use the capital to expand its research team and push its judgmental forecasting product into more enterprise and government workflows.
Betting On 'Superhuman' Judgmental Forecasting
Mantic adapts existing foundation models with reinforcement learning and human-in-the-loop pipelines so the system can answer messy, real-world questions weeks or months into the future — from geopolitical events to product launches and market reactions. In the fall 2025 Metaculus Cup, Mantic's bot took first place, edging past a field of top human forecasters and marking one of the first public benchmarks where an AI beat professional predictors on out-of-distribution questions.
A Familiar Cap Table For AI Bets
Radical Ventures, the Toronto-based AI-focused firm that has backed companies including Cohere and Waabi, is leading the round; Microsoft's M12 arm and London-based Balderton Capital join. Notably, Thinking Machines Lab — the AI lab founded in 2025 by former OpenAI CTO Mira Murati that itself commands a multi-billion valuation — is also in as a strategic investor, an unusually direct bet from one AI lab on another. The seed round follows a period in which VCs have poured record capital into applied AI research shops with defensible model IP.
Enterprise, Policy And Intelligence Buyers
Mantic sells its forecasting system to hedge funds, corporate strategy teams, defense customers and government agencies that need calibrated probability estimates rather than freeform LLM opinions. The startup joins a small but growing cohort of "prediction market"-adjacent AI companies including Metaculus and Manifold, but is one of the few positioning itself as a pure-play, enterprise-grade forecasting engine. Reporting based on coverage from Tech Startups, Yahoo Finance and Crypto Briefing.