Olas Released New AI Forecasting Model
The 14-billion parameter model achieved 75.8% accuracy on prediction-market tests.
Updated on Sept. 22, 2026 in Artificial Intelligence

Live Poll
Do you trust specialized, smaller AI models to perform as reliably as large general-purpose systems?
Olas has launched the Olas-Predict-R1-14B, an open-weight forecasting model fine-tuned on over 214,000 prediction records. The model outperformed its base architecture by delivering 75.8% accuracy on test markets.
Why it matters
By leveraging thousands of resolved market outcomes, the model offers a specialized alternative to general-purpose AI for probability estimation. Its ability to run on a single GPU increases accessibility for developers seeking to automate complex forecasting tasks.
The 14-billion parameter model is based on DeepSeek-R1-Distill-Qwen-14B and utilizes 214,529 records from 5,116 resolved markets. It demonstrates a 20% relative reduction in forecasting error compared to the base model.
The players
Olas
Olas is a decentralized autonomous organization that builds infrastructure for AI agents and prediction markets.
Olas Predict
Olas Predict is an operational forecasting platform that compiles market data and verified real-world outcomes.
The details
The model processes external data to generate probability estimates for event outcomes based on historical patterns. Developers designed it to function on single-GPU hardware, aiming to provide efficient performance compared to larger, general-purpose AI systems.
Timeline
Olas Predict has operated since 2023.
The new forecasting model was released on September 22, 2026.
The Tech Race
The release of Olas-Predict-R1-14B builds upon the DeepSeek-R1-Distill-Qwen-14B foundation to address the need for specialized prediction tools. This advancement reflects a broader industry shift toward optimizing smaller, task-specific models that outperform general-purpose counterparts in efficiency.
Developers and researchers can utilize the model's single-GPU capability to deploy automated forecasting agents without significant infrastructure costs. The open-weight nature of the software allows for broader integration into custom analytical workflows and predictive applications.
The takeaway
The development demonstrates that fine-tuning base models with large, curated datasets can significantly boost predictive accuracy for specific industry needs. Organizations interested in automation should consider how specialized models can improve decision-making speed compared to off-the-shelf solutions.
Further reading
For more developments in this field, explore our latest updates on Artificial Intelligence.
Source note: This article includes information reported by IT Brief Australia.
Live Poll
Do you trust specialized, smaller AI models to perform as reliably as large general-purpose systems?



