from the AI 2025 Forecasting Survey by AI Digest

The best performance by an AI system on Re-Bench as of December 31st 2025.

This resolution will use AI Digest as its source. If the number reported is exactly on the boundary (eg. 0.80) then the higher choice will be used (ie. 0.8-1.0). 

Any AI system counts if it operates within realistic deployment constraints and doesn't have unfair advantages over human baseliners.

Tool assistance, scaffolding, and any other inference-time elicitation techniques are permitted as long as:

 elicitation technique (which automatically grades and chooses the best out of k outputs from a model) is a common example that we 

 because human baseliners in RE-Bench also have access to scoring metrics (e.g. loss/runtime). So PASS@k doesn't constitute a clear unfair advantage.

If there is evidence of training contamination leading to substantially increased performance, scores will be accordingly adjusted or disqualified.