Summary
According to a post on X by @autonolas, Valory trained Olas-Predict-Qwen3.8-27B, an open-weight forecasting model built on Qwen3.8-27B. In Valory's own evaluation, it reached 67.3% balanced accuracy against 68.5% for GPT-5.6 across 6,824 prediction markets created after GPT-5.6's training cutoff. Valory says most of the gain came from the stronger base model. These figures come from Valory's own evaluation and have not been independently checked by VQV.
AI-assisted summary based on the listed source.
What happened
Olas-Predict-Qwen3.8-27B, built on Qwen3.8-27B, scored 67.3% balanced accuracy against GPT-5.6's 68.5% on 6,824 prediction markets created after GPT-5.6's training cutoff, per Valory's own evaluation. Valory says most of the gain came from the stronger base model. Why it matters: A small open model nearly matching...
Why it matters
For teams building forecasting tools, the post suggests a small open model can come close to a frontier model on this task, which could lower the cost of prediction tools. Developers weighing open-weight models against closed frontier models may want to see whether independent evaluations reproduce the close result. Until then, the comparison rests on Valory's own test.
Signal Intelligence
Signal Strength 82%
Technical label WATCH
Public Interest 34
Category OPEN SOURCE
Reader Depth TECHNICAL
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
Recognizable Entity Score 0
Practical Impact Score 64
Novelty Interest Score 70
Consequence Score 10
Curiosity Score 0
Shareability Score 53
Why this is here
VQV surfaced this signal because it is recent, relevant to Open Source LLMs, connected to Scobleizer News Pulse (X).
