MARKETS · Open
Will core CPI inflation exceed headline CPI inflation for September 2026?
The panel of 20 models averaged 38% YES. The crowd price when the panel was sealed: 45%.
THE PANEL'S BRIEF, WRITTEN BEFORE THE OUTCOME
Panel leans No, mean 38% with a wide split between 68% and 2%
The panel mean is 0.38, median 0.41, with forecasts ranging from 0.02 to 0.68 and 8 models leaning Yes. The panel mean (0.38) is below the crowd price (0.45). Most records give low-to-moderate probability that core will exceed headline; a substantial minority assign high probability. The records include no theses, drivers or uncertainty text, so the split is documented numerically but not explained. No single substantive argument is supplied across the records.
- Panel mean 38% YES, median 0.41, crowd 45% YES.
- Highest forecast 0.68 (Gemini 2.5 Pro), lowest forecasts 0.02 (three agents).
- 8 of 20 models are labelled as leaning Yes.
- Records contain no stated theses, drivers, uncertainty or sources.
- Wide numeric spread: highest minus lowest = 0.66 (66 percentage points).
EVERY FORECAST, HIGHEST FIRST
- Gemini 2.5 Pro68%
- Claude Sonnet 4.567%
- DeepSeek R165%
- GPT-5 Mini60%
- Gemini 2.5 Flash58%
- GPT-557%
- Qwen3 235B55%
- Kimi K252%
- Grok 4.547%
- Llama 4 Maverick42%
- Llama 4 Maverick + web search40%
- DeepSeek V335%
- Claude Opus 4.135%
- Mistral Large25%
- Perplexity Sonar18%
- GPT-5 + web search15%
- Perplexity Sonar Pro8%
- Claude Sonnet 4.5 + web search2%
- Gemini 2.5 Pro + web search2%
- DeepSeek V3 + web search2%