Valory unveils open-weight Olas-Predict-Qwen3.8-27B, scoring 67.3% balanced accuracy on prediction markets

Valory has unveiled Olas-Predict-Qwen3.8-27B, an open-weight forecasting model that the company says comes within 1.2 percentage points of its GPT-5.6 comparator on a test of real-world prediction markets. The announcement was reported by Unite.AI.

Key figures

• Model: Olas-Predict-Qwen3.8-27B
• Parameters: 27 billion
• Weights: open
• Valory's model score: 67.3% balanced accuracy
• GPT-5.6 score: 68.5% balanced accuracy
• Test conditions: both systems received the same market evidence

The scores come from Valory's internal evaluation, so they are company claims and have not been independently verified. Pricing, context window and training cutoff date have not been published in the available summary.

What this means

According to Valory's own evaluation, a 27-billion-parameter specialist with open weights scored 67.3% balanced accuracy against 68.5% for GPT-5.6 on the same market evidence. Unite.AI frames the result as raising a practical question about how much of a specialist task's performance a smaller dedicated model can capture. Because the comparison is internal, teams considering the model for forecasting work will want to check it on their own market data.