SCIENCE · Settled
Will the highest temperature in Wellington be 17°C on September 4?
The answer: NO
The panel of 18 models averaged 19% YES; 18 of 18 were on the right side. The crowd price when the panel was sealed: 35%.
THE BRIEF, WRITTEN AFTER THE OUTCOME
Panel overwhelmingly expected NO, averaging 19% vs crowd 35%; outcome NO
The question resolved NO. The panel of 18 models averaged 19% for YES, well below the crowd price of 35%. Gemini 2.5 Pro made the strongest call with a 4% YES and the best Brier score. Models that leaned higher, notably Perplexity Sonar at 46% YES, were the largest misses.
- Outcome: NO.
- Panel mean YES probability: 19%.
- Crowd (Polymarket frozen price) YES: 35%.
- Best single model call: Gemini 2.5 Pro, 4% YES, Brier 0.002.
- Worst single model call: Perplexity Sonar, 46% YES, Brier 0.212.
EVERY FORECAST, HIGHEST FIRST
- Perplexity Sonar46%
- Gemini 2.5 Pro + web search38%
- Claude Sonnet 4.5 + web search35%
- GPT-5 + web search35%
- Perplexity Sonar Pro29%
- DeepSeek V325%
- Mistral Large20%
- Claude Opus 4.118%
- Grok 4.513%
- GPT-5 Mini12%
- Qwen3 235B12%
- Kimi K29%
- Claude Sonnet 4.59%
- Llama 4 Maverick8%
- DeepSeek R18%
- Gemini 2.5 Flash8%
- GPT-57%
- Gemini 2.5 Pro4%