Analysis
Mantic, a London-based AI startup building forecasting systems, raised a $25 million seed round led by Radical Ventures, with Balderton Capital, Microsoft's M12 and Mira Murati's Thinking Machines Lab also participating, according to TechStartups and Yahoo Finance. No valuation was disclosed.
What Mantic Actually Does
Mantic was co-founded by Toby Shevlane, a former Google DeepMind researcher, alongside Ben Day. The company adapts frontier language models to assign calibrated probabilities to future political, economic, business and cultural events -- a narrower and more measurable product than general-purpose chat or coding assistants, since forecasting accuracy against real-world outcomes is the entire value proposition rather than a supporting benchmark. Mantic drew industry attention after its system performed competitively against experienced human forecasters in the Summer 2026 Metaculus Cup, an online tournament where AI systems have increasingly matched or exceeded top human competitors on some categories of questions.
“## What Mantic Actually Does Mantic was co-founded by Toby Shevlane, a former Google DeepMind researcher, alongside Ben Day.”
The Competitive Landscape
Mantic sits adjacent to, rather than in direct competition with, Metaculus itself -- Metaculus is the tournament platform and community forecasting hub, while Mantic is building a standalone commercial forecasting product. The broader category includes prediction markets like Polymarket and Kalshi, which price events through trading rather than model inference, and Good Judgment, the longer-running human-forecaster consultancy built on the original IARPA-funded superforecasting research. Mantic's pitch is that a model-based system can now compete with the best of those human-driven approaches on speed and coverage, even if it hasn't yet displaced them on trust for the highest-stakes questions.
Why This Investor Roster Matters
Radical Ventures, Balderton and M12 are multi-stage funds that don't typically anchor $25 million seed rounds without strong conviction, and Thinking Machines Lab's participation -- Mira Murati's company, itself one of the most closely watched new AI labs -- adds a research-credibility signal distinct from a pure financial bet. That combination suggests investors are underwriting the Metaculus result as a genuine capability proof point rather than treating Mantic as a standard early-stage bet on a compelling deck.
The Numbers In Context
$25 million is a modest round by this week's standards -- Nscale's IPO filing alone discloses over $100 billion in contracted value, and Crusoe just closed at a $30.9 billion valuation -- but forecasting-as-a-service is a different category entirely, closer to a data and analytics product than infrastructure, and doesn't require the capital intensity of compute-heavy AI bets.
The risk is durability of the edge: tournament performance in a controlled competition like the Metaculus Cup doesn't automatically translate into a defensible commercial product, and forecasting accuracy claims are notoriously difficult to verify outside of exactly this kind of head-to-head, scored competition. Whether enterprises and institutions are willing to pay for AI-generated probability estimates on decisions that matter -- rather than treating forecasting as a research curiosity -- is the open question the $25 million is meant to answer.