Codename: HORIZON

Resolves YES if Model Evaluation & Threat Research (METR) publishes an 80%-success task horizon of at least 16 hours for any AI model publicly released before January 1, 2028.

This is METR’s central estimate for p80_horizon_length.estimate; the threshold is 960 minutes. The model must be released before the deadline, but its measurement may arrive later.

The qualifying measurement must be published by April 30, 2028. If METR stops publishing comparable results, MNX will use the closest credible equivalent.

Keep an eye on mnx.fi 👀