Mantic Raises $25 Million After AI Forecasting Breakthrough
Artificial intelligence startup Mantic has raised $25 million in seed funding, putting fresh investor money behind an ambitious effort to build AI systems capable of forecasting real-world events more accurately than expert human forecasters.
The London-based company attracted attention after its system outperformed every human participant in the Summer 2026 Metaculus Cup, an online forecasting tournament covering political, economic, technological and cultural events.
The funding round was led by Radical Ventures, with participation from Microsoft's M12 venture fund, Thinking Machines Lab, Balderton Capital and other investors. The valuation of the company was not disclosed.
The funding comes as AI companies increasingly explore applications beyond generating text, images and software code. Mantic is targeting a different problem: using AI to estimate the probability of uncertain future events.
Mantic Beat All 676 Human Participants — But One AI Finished Higher
Mantic says its system finished ahead of all 676 human participants, including professional forecasters, in the Summer 2026 Metaculus Cup.
That result requires an important qualification.
Mantic did not finish ahead of every participant overall. Reuters reported that it beat every human contestant but finished behind one AI bot called laertes. The broader performance of AI systems in the competition marked the first time technology had dominated the tournament.
The result therefore provides evidence of Mantic's performance on a specific set of forecasting questions under the tournament's scoring rules. It should not be interpreted as proof that the system can predict every kind of future event better than humans.
What Does Mantic Actually Predict?
Unlike conventional forecasting systems built primarily around numerical or historical datasets, Mantic is designed for events that also require reasoning and judgement.
The company says its system focuses on medium-term predictions, generally ranging from one week to one year, across areas including geopolitics, business, policy, technology and culture.
Instead of simply asking an AI model what will happen, Mantic specialises frontier AI models developed by other laboratories for forecasting.
According to CEO and co-founder Toby Shevlane, the company tests its predictive systems against historical events, evaluates their performance and then uses those results to improve forecasting ability.
“We're now upgrading the level at which humans can understand the future,” Shevlane told Reuters.
From Google DeepMind to an AI Forecasting Startup
Mantic was co-founded in 2024 by Toby Shevlane and Ben Day.
Shevlane previously worked as a research scientist at Google DeepMind. He told Reuters that the idea behind Mantic partly emerged from his need to anticipate global developments relevant to artificial intelligence while working there.
The startup is attempting to turn probabilistic forecasting into a scalable AI product that organisations can use when making decisions involving uncertain future events.
Rather than providing a simple yes-or-no prediction, forecasting systems typically assign probabilities to possible outcomes. Those forecasts can then be evaluated after the event occurs.
How Mantic Avoided Following the Crowd
One of Mantic's claimed advantages during the tournament was its ability to avoid what Shevlane described as herd mentality.
Reuters highlighted an example involving Shakira's songs. Human forecasters overwhelmingly expected “Dai Dai” not to overtake “Waka Waka” on the Billboard Hot 100. That consensus proved incorrect, while Mantic had not placed as much confidence in the prevailing view.
Another example came from Colombia's presidential election.
According to Shevlane, Mantic gave Abelardo De La Espriella roughly a 40% probability of winning at a point when the consensus forecast was around 30%. De La Espriella ultimately won.
These examples illustrate an important characteristic of probabilistic forecasting: success is not simply about declaring one outcome certain. Systems are rewarded for assigning well-calibrated probabilities across many questions.
Why Investors See Commercial Potential
The significance of AI forecasting extends beyond forecasting competitions.
Businesses routinely make decisions based on uncertain events—from interest rates and elections to regulation, supply chains, product launches, geopolitical developments and market conditions.
Mantic says hedge funds are already using its technology to predict market-moving developments across areas including geopolitics, macroeconomics and business. The company also says Fortune 500 customers are adopting the system for strategic decisions including mergers and acquisitions and product releases. These customer claims come from Mantic itself, and the company has not publicly identified those customers.
Radical Ventures partner Aaron Rosenberg, who has joined Mantic's board, told Reuters that companies and government agencies around the world have expressed interest in the technology, with some already integrating it. Mantic declined to disclose customer names.
Hedge funds and trading firms are particularly interested because improvements in forecasting market-moving events could potentially have direct financial value.
“If Mantic is superhuman as it is, they can make money off of that immediately,” Rosenberg told Reuters.
What Mantic Plans to Do With the $25 Million
Mantic says the new capital will be used to significantly expand its team as well as the computing resources and data supporting its forecasting technology.
The round was led by Radical Ventures, with Mantic listing Balderton, Thinking Machines, DRW, FT Ventures, M12, Episode 1, Charlie Songhurst and Thomas Wolf among the participants.
The funding gives the startup additional resources to test whether its strong tournament performance can translate into consistently useful forecasts for businesses, investors and governments.
Why Mantic's Result Matters for AI
Much of the recent AI boom has focused on systems that generate content or perform tasks based on existing information.
Forecasting poses a different challenge.
An AI forecaster must gather current information, reason about uncertain developments, assign probabilities and then be judged against what actually happens.
That makes forecasting tournaments particularly useful benchmarks because predictions can be recorded before an event and objectively scored afterward.
Mantic's Metaculus performance suggests specialised AI forecasting systems are becoming increasingly competitive with skilled humans on such benchmarks. But tournament success and dependable real-world decision-making are not identical: performance still depends on the questions, information available, scoring methodology and forecasting horizon.
The next test for Mantic will therefore be commercial as much as technical—whether the accuracy demonstrated in structured forecasting competitions can produce measurable advantages when organisations use its predictions to make consequential real-world decisions.
