EXCLUSIVE: Small AI Model Matches GPT-4.1 at Forecasting as Olas Founding Member Says Frontier Models Face ‘Increasing Economic Pressure’

Olas unveiled a new AI model Tuesday that was trained on more than 200,000 prediction market forecasts and matched OpenAI’s GPT-4.1 in a test of its ability to predict real-world events. Olas-Predict-R1-14B achieved 75.8% accuracy across 2,628 previously unseen markets, compared with 75.4% for GPT-4.1 and 71.4% for the underlying DeepSeek model, according to benchmark materials shared with Benzinga. David Minarsch, CEO of Valory and founding member of Olas, told Benzinga the results suggest cheaper, specialized AI models are putting “increasing economic pressure” on the industry’s most advanced general-purpose systems. • NVIDIA stock is trading near recent highs. What should traders watch with NVDA? Prediction Markets Become AI Training Data Olas, which develops autonomous AI agents that research and trade on prediction markets, trained the model using 214,529 forecasts from 5,116 resolved markets. Each forecast eventually gets an answer, letting researchers see which predictions were right and which were too confident. Minarsch said the models were not shown existing market odds during the benchmark, meaning they could not simply copy the crowd. Fine-tuning also improved the model’s Brier score by roughly 20%. The metric rewards accurate probabilities and penalizes confident mistakes. “The headline in AI has been that better performance requires bigger, more powerful general-purpose models,” Minarsch said. “Our results point to another path.” Smaller Models Pressure AI Economics Minarsch said frontier models face “increasing economic pressure from such specialized models” and may need to find new markets to maintain their growth rates. That comes as Big Tech pours record sums into AI infrastructure. TrendForce expects the nine largest cloud providers to spend more than $886.7 billion in total capital expenditures in 2026, up roughly 90%, with AI infrastructure driving much of the increase. NVIDIA Corp. (NASDAQ:NVDA) sits at the center of that spending boom because larger models generally require more GPUs and data-center capacity. If smaller specialist models can handle more workloads with less computing power, that could change the economics behind some of that demand. Minarsch stopped short of saying the spending is misplaced. Olas-Predict itself builds on a model that required far more computing power to create, but he argued cheaper specialist systems can pressure closed-model economics by handling particular jobs just as well. Olas says its model can run on a single GPU and its weights are publicly available, letting developers run it themselves instead of paying a closed AI provider for each forecast. See More: Top Momentum Stocks Thomson Reuters Bets on Specialist AI Thomson Reuters Corp. (NASDAQ:TRI), which sells legal, tax and financial information and software, is making a similar bet. The company invested about $40 million to build a specialist model using an open foundation model and proprietary data, saying ownership gives it greater control over performance, privacy and costs. In an experiment requested by Benzinga, Olas-Predict gave Nvidia a 70% chance of ending 2026 as the world’s largest company by market capitalization. Polymarket traders currently put Nvidia at 78%. Olas says the model is available for download on Hugging Face, while the Olas marketplace offers pay-per-call access. Image: Shutterstock Read Also: China Slams Brakes on Humanoid Robot IPOs After Unitree’s 55% Plunge — Tesla Optimus Faces Commercialization Test