Will Google have the best AI model at the end of October 2026?

THE PANEL'S BRIEF, WRITTEN BEFORE THE OUTCOME

Panel low at 28%, wide split driven by a 72% high and 10% lows, far below crowd 73%

The panel mean is 28%, median 27%, with a 62 point spread between the highest (72%) and lowest (10%) forecasts. Most records cluster between 25% and 35%, signalling general scepticism about Google being top by the check date. The panel splits because one agent, Perplexity Sonar Pro, gives 72% while several agents give 10%, and none of those records supply theses or drivers. The single item doing the most work is Perplexity Sonar Pro's 72% yes probability, which dominates the high side of the distribution.

• Panel mean: 28%, median: 27%, highest: 72%, lowest: 10%.
• Crowd price frozen at 73%, panel sits far below that.
• Most models give between 25% and 35%, indicating broad scepticism.
• A single high forecast (Perplexity Sonar Pro 72%) creates the large spread.
• Several models state 10% with no supporting drivers in their records.