In 2025, DeepMind announced that AlphaEvolve helped reduce Gemini’s training time by 1%. This market asks whether AlphaEvolve will deliver a 5% or greater reduction in training time for Gemini models by the end of 2025.

This market resolves YES if, by 31 December 2025, DeepMind or Google publicly state that AlphaEvolve has contributed to a 5% or greater reduction in the training time, compute usage, or energy cost of any full-scale Gemini model (e.g. Gemini 2.5 or successors).

The improvement must:

• Be attributed specifically to AlphaEvolve

• Apply to training

• Apply to large-scale Gemini models

The market resolves NO if no such claim is made by 31 December 2025, or if only improvements below 5% are publicly reported.

I will not bet in this market. Please let me know if something is unclear.