AI Forecasting Startup Mantic Raises $25M in New Funding

London-based AI company Mantic has raised $25 million in seed funding after its forecasting system beat every human competitor in a major prediction tournament. The round, announced on September 18, 2026, was led by Radical Ventures, with participation from Microsoft’s M12, Thinking Machines Lab, Balderton Capital, DRW, FT Ventures, Episode 1, and angels including Charlie Songhurst and Thomas Wolf. Radical partner Aaron Rosenberg has joined the board. The company had previously raised about $4 million in a pre-seed round.

A Public Test of “Superhuman” Forecasting

Mantic’s calling card is its result in the summer 2026 Metaculus Cup, a tournament in which hundreds of people assign probabilities to political, economic, technological, and cultural outcomes. The company says its system outperformed all 676 human participants, including professional forecasters. It also beat other AI startups and frontier models run on a standard competition harness. Only one other bot, called laertes, finished ahead of it. Company co-founders say experts had recently assumed this kind of result was still a decade away.

The tournament included questions that reward research and judgment rather than clean historical data. Mantic has cited examples such as a higher-than-consensus probability that Abelardo De La Espriella would win Colombia’s presidential election, skepticism about a consensus view on a Shakira chart ranking, Spider-Man box-office estimates, and the timing of events such as a possible reopening of the Strait of Hormuz. Reuters noted that AI systems dominated the field for the first time. A single contest does not prove a system can forecast every kind of event, but it gave investors a rare public scoreboard.

What Mantic Builds

Mantic focuses on judgmental forecasting: messy, medium-term questions about business, geopolitics, policy, technology, and culture. It does not claim to replace weather models or high-frequency market signals. Instead, it specializes frontier models from other labs for prediction, tests them on historical questions, scores the results, and iterates. CEO Toby Shevlane, a former Google DeepMind researcher who co-founded the company in 2024 with Ben Day, has described the goal as raising the level at which people can understand the future. The team includes alumni from DeepMind, Cambridge, Oxford, and trading firms such as Citadel.

Who Wants Better Odds

The commercial pitch is straightforward. Hedge funds, Fortune 500 companies, and government agencies want better probabilities for mergers, product launches, supply-chain risk, demand, and political shocks. Mantic says some leading hedge funds already use the system in trading decisions and that Fortune 500 customers are using it for strategy. Rosenberg said companies and agencies around the world have shown interest, and that some have already integrated the product. Named customer lists remain limited in public reporting, which is typical at this stage but leaves the depth of revenue still hard to judge from outside.

Where the Money Goes Next

Mantic plans to use the new capital to hire across research, engineering, product, sales, and operations, and to buy more compute and data. The company frames 2026 as the start of an era in which machines can outperform top human generalist forecasters on open-ended questions. That claim will be tested less in press releases than in the next Metaculus Cups, live customer decisions, and whether better probabilities actually change outcomes when money and policy are on the line. For now, a tournament win and a well-known investor syndicate have turned a niche AI problem into one of the clearer early bets on automated judgment.

Source: Reuters