AI · Settled
Will Bytedance have the best AI model at the end of August 2026?
The answer: NO
The panel of 13 models averaged 12% YES; 13 of 13 were on the right side. The crowd price when the panel was sealed: 0%.
THE BRIEF, WRITTEN AFTER THE OUTCOME
Outcome NO; panel strongly favoured NO and low‑prob models scored best
The question resolved NO. The crowd price was frozen at NO, p_yes 0. The 13‑model panel had a mean p_yes 12% and median 9%, all models called NO. Gemini 2.5 Pro and GPT-5 were best, each giving 2% to YES. The largest misses were models that assigned the highest YES probabilities, up to 30%.
- Outcome: NO, resolved 2026-09-01 04:00:01.
- The crowd was the Polymarket frozen price, p_yes 0, it called NO.
- Panel of 13 models, all called NO; mean p_yes 12%, median 9%.
- Best numerical calls were 2% YES from Gemini 2.5 Pro and GPT-5.
- Worst single call was Mistral Large at 30% YES, highest Brier 0.09.
EVERY FORECAST, HIGHEST FIRST
- Mistral Large30%
- Llama 4 Maverick25%
- Qwen3 235B18%
- DeepSeek V315%
- Claude Opus 4.112%
- DeepSeek R112%
- Grok 4.59%
- Gemini 2.5 Flash8%
- Kimi K28%
- GPT-5 Mini7%
- Claude Sonnet 4.56%
- GPT-52%
- Gemini 2.5 Pro2%