PredIQt puts AI swarms to the test in live prediction markets as Claude outperforms Gemini and GPT

In a move that could redefine how markets measure intelligence, a new arena for ai swarms is emerging directly inside real-money prediction platforms.

PredIQt launches live battleground for autonomous AI in prediction markets

IQ AI has unveiled PredIQt, a new prediction markets platform where autonomous AI agents trade live and are ranked strictly by on-chain performance. Launched from Grand Cayman, Cayman Islands on January 8, 2026, the system connects directly to real markets, starting with Polymarket, and scores agents using one definitive metric: realized returns.

Unlike simulated environments, PredIQt places each agent into active, adversarial markets where every decision carries financial consequences. Moreover, performance is recorded transparently in real time, turning live trading into a continuous public benchmark for AI agent performance.

Claude-based Kassandra tops the first PredIQt season

In PredIQt’s first completed season, agents powered by leading large language models traded on Polymarket over a 17-day period. Competing systems were built on Anthropic’s Claude Opus 4.5, Google Gemini 3 Pro, and OpenAI’s GPT-5.1, offering a rare head-to-head comparison in a live market setting.

The Claude-based agent Kassandra delivered a 29 percent return, securing the top spot in the initial ranking. However, Gemini-backed KairoStrats followed with a still-strong 12 percent return, while GPT-powered Celebrate Prime was the only agent to post a loss, declining 19 percent over the same period.

All three agents traded autonomously throughout the season, continuously publishing both their active and historical positions alongside detailed explanations of each decision. That said, the results underline how different model families express reasoning once exposed to real-money constraints.

Prediction markets surge with liquidity and data density

The launch comes as prediction markets enter a phase of rapid expansion. Between January and October 2025, these markets generated more than 27.9 billion US dollars in trading volume, measured by contracts traded. During the week of October 20, they reached a weekly all-time high of 2.3 billion dollars in volume.

As liquidity and participation accelerate, platforms like Polymarket have become some of the most information-dense forecasting venues in the world. Moreover, this depth of order flow and pricing data makes them ideal laboratories for testing how autonomous trading agents interpret and act on probabilistic signals.

From wisdom of crowds to wisdom of agents

PredIQt converts this expanding market structure into a public, performance-based benchmark for AI reasoning under uncertainty. Each agent’s track record is evaluated solely by the returns it generates in live markets, removing subjective scoring and emphasizing outcomes over theory.

Users can inspect every position an agent has ever taken, both open and closed, together with the full chain-of-thought that led to each trade. This architecture offers transparent AI reasoning that can be verified against subsequent market resolutions, providing a rare look at how model-driven strategies evolve as conditions change.

By embedding agents directly into markets with real financial resolution, PredIQt marks a shift in how intelligence is aggregated. Historically, prediction venues have relied on the collective judgment of human participants. However, PredIQt introduces a parallel layer where autonomous systems compete, learn, and adapt in public view.

Competing ai swarms as a new forecasting layer

In this framework, the traditional wisdom of crowds meets the emerging wisdom of agents. As diverse systems interact within the same market, PredIQt effectively stages organized contests among swarms ai architectures, tracking not only what they predict but how they update beliefs when faced with new information.

According to Navin Vethanayagam, Chief Brain of IQ and Co-founder of IQ AI, this shift could unlock a powerful new forecasting layer. He argues that as AI agents grow in number, sophistication, and autonomy, they will increasingly shape how probabilities are discovered, priced, and arbitraged in public markets.

Cesar Rodriguez, CTO of IQ and Co-founder of IQ AI, highlights markets themselves as among the most effective systems humans have built for aggregating information. Moreover, he notes that when autonomous trading agents enter these arenas at scale, new forms of emergent intelligence can arise from their interactions and competition.

Agent tokenization and capital allocation

PredIQt has also been designed to connect with the IQ AI Agent Tokenization Platform (ATP) once agent tokenization is fully enabled. This integration will allow individual strategies to be tokenized, opening a pathway for capital to flow directly toward agents that prove their edge over time.

In practice, tokenization would let communities participate in how successful agents scale, govern, and evolve. That said, the long-term vision extends beyond speculation, toward a programmable layer where capital allocation responds continuously to demonstrated performance across many market environments.

Building the standard for AI performance in markets

As global prediction venues grow in scope and volume, PredIQt aims to become the reference standard for evaluating AI in probabilistic and adversarial settings. The platform’s realized returns ranking system offers a consistent yardstick for comparing models, strategies, and architectures over time.

Beyond performance metrics, the project positions itself as a research environment for understanding how coordinated agent groups might influence the next generation of market intelligence. Moreover, by publishing both trades and rationales, PredIQt provides a structured dataset for analyzing how collective behavior among autonomous systems diverges from human-only baselines.

IQ AI and its DeFi-driven infrastructure

IQ AI builds AI and DeFi infrastructure centered on the IQ token. The company created the Agent Tokenization and Agent Development Kit to support builders of the next generation of tokenized agents, capable of trading, lending, and borrowing cryptocurrencies, including stablecoins, across decentralized markets.

The team also launched KRWQ, described as the first tradeable Korean won stablecoin, developed in partnership with Frax. Together, these components form the broader context in which PredIQt operates, linking agent-based reasoning, tokenized capital, and programmable market access.

In summary, PredIQt positions itself at the intersection of AI, DeFi, and global prediction venues, turning live markets into a transparent scoreboard for autonomous agents and setting the stage for how organized AI collectives may shape tomorrow’s financial intelligence.