Next Mythos-Class Model released by September 1, 2026?

THE BRIEF, WRITTEN AFTER THE OUTCOME

Panel mostly lowballed a YES; one model nailed it with 70%

The event resolved YES. The crowd priced it at 38% while the panel mean was 15% and the median 4%. DeepSeek V3 was the lone strong YES at 70% and was the only panel model to call it. Most misses clustered at very low probabilities, with several models at 2%, 4%. The uninformed prior control (50%) also called it YES.

• Outcome: YES.
• Crowd price frozen at 0.38, did not call it.
• Panel mean_p_yes 0.15, median_p_yes 0.04, only 1 of 13 models called YES.
• DeepSeek V3 gave p_yes 0.7 and was the only correct panel caller.
• Many models assigned p_yes 0.02, 0.04 and hence were badly wrong.