Mantic Raises $25 Million After AI Forecasting Win

Mantic has raised $25 million after its AI forecasting system outperformed every human entrant in a major prediction tournament. The result gives the London startup evidence for a difficult claim: that specialized AI can improve decisions about uncertain political, economic and business events, not merely summarize what has already happened.

A Tournament Result Gives the Funding Story Weight

The seed round was reported by Reuters and confirmed in Mantic’s announcement. Radical Ventures led the financing at an undisclosed valuation, with participation from Microsoft’s M12, Thinking Machines Lab, Balderton Capital and other investors.

The timing matters because Mantic can point to measured performance rather than a controlled demonstration. In the summer 2026 Metaculus Cup, its system beat all 676 human participants and finished behind only one other bot. Forecasts covered political, economic and cultural outcomes, with competitors scored on the probabilities they assigned before events were resolved.

Mantic Specializes Frontier Models for Prediction

Mantic does not present itself as another general-purpose model laboratory. Its product adapts frontier models from other developers for probabilistic forecasting. The system tests predictions against historical data, grades performance and uses those results to improve future estimates. That narrower design supports an argument examined in Business 2.0’s comparison of specialized research agents and general-purpose AI: workflow-specific evaluation can matter more than broad benchmark strength.

Co-founders Toby Shevlane and Ben Day formed the company in 2024. Shevlane previously worked at Google DeepMind, where forecasting global developments relevant to AI helped inspire the idea. Mantic’s launch statement frames the product around geopolitics, policy, business, technology and culture—areas where evidence changes quickly and confident narratives often conceal uncertainty.

Investors Are Backing Decision Infrastructure

Lead investor Radical Ventures focuses on AI companies, while M12 brings a strategic connection to Microsoft’s enterprise ecosystem. The syndicate also includes Thinking Machines Lab and European venture firm Balderton Capital. Mantic says the capital will expand its team and the compute and data supporting its forecasts.

This is a bet on decision infrastructure rather than consumer engagement. Financial traders could monitor market-moving events; executives could assess regulatory or competitive risks; governments could update scenarios as new evidence arrives. The opportunity resembles the enterprise-agent market behind Vals AI’s real-world agent testing, but Mantic’s output is a calibrated probability rather than an autonomous action.

Winning a Competition Is Not the Same as Winning Trust

Forecasting systems face a different burden from chatbots. A plausible explanation is insufficient if the probability is poorly calibrated or the source evidence is weak. Customers will need historical performance by question type, records of forecast revisions and clear separation between model confidence and business impact. Unusual events can also break patterns that looked reliable in back-testing.

Mantic must therefore prove that tournament performance transfers into operational settings where questions are ambiguous and incentives differ. Financial customers will compare its signals with experienced analysts, including the judgment challenges explored in Business 2.0’s coverage of market-crash forecasting. Governments will require auditability, security and explicit accountability for decisions.

The Commercial Test Starts After the Benchmark

The $25 million round gives Mantic resources to move from technical validation into customer deployment. It also raises expectations. Buyers will ask whether forecasts improve decisions after costs, false confidence and integration work are included—not merely whether the system ranks above individuals in a tournament.

If Mantic can demonstrate that advantage, forecasting could become a distinct enterprise AI category spanning finance, policy and corporate strategy. The funding follows investor demand for defensible AI applications seen in Discovery Loop’s AI financing and wider market appetite discussed around Anthropic’s reported IPO preparations. Mantic now has to turn a striking competitive result into repeatable institutional value.