TL;DR
The crowd treats a positive July core CPI print as near-certain (96 cents above 0.0%). The models were handed the actual recent prints — including June's slightly negative core reading — plus July's 30% oil rally, and asked to price the whole ladder coherently. The blend puts "above 0.0%" at 91%, and the interesting disagreements live in the middle rungs.
Each model produced one distribution across all six thresholds in a single pass, so every model column is internally consistent — the same whole-ladder protocol our backtesting found far more accurate than pricing rungs independently. This market settles on the BLS release in mid-August, making it one of the fastest publish-to-graded loops on our scoreboard.
The Ladder: July Core CPI Month-Over-Month
| Print above | Kalshi | AI blend | ChatGPT (GPT-5.5) | Claude Fable | Claude Opus | Claude Sonnet | Gemini 3.1 Pro | GLM 5.2 | Kimi K3 | DeepSeek V4 |
|---|---|---|---|---|---|---|---|---|---|---|
| 0.0% | 96¢ | 91% | 94% | 93% | 93% | 80% | 95% | 88% | 93% | 93% |
| 0.1% | 77¢ | 77% | 83% | 80% | 79% | 62% | 82% | 74% | 78% | 78% |
| 0.2% | 11¢ | 51% | 58% | 52% | 46% | 42% | 52% | 48% | 54% | 53% |
| 0.3% | 22¢ | 24% | 29% | 22% | 18% | 24% | 22% | 22% | 27% | 26% |
| 0.4% | 2¢ | 8% | 10% | 7% | 6% | 10% | 7% | 9% | 9% | 9% |
| 0.5% | 6¢ | 2% | 3% | 2% | 2% | 3% | 2% | 3% | 2% | 2% |
A note on the Kalshi column: those are last-trade prices, and on thin rungs the last trades can be stale or out of order (the 0.2% and 0.3% rungs above traded out of sequence). The model columns are each a single coherent distribution.
The Reasoning
ChatGPT (GPT-5.5): The recent base rate points to a core CPI process centered in the low-to-mid 0.2s: January through May averaged about 0.26% MoM, while June’s -0.02% was a sharp downside outlier relative to the preceding trend. That negative print raises the odds of some payback in July if it reflected temporary discounts, residual seasonality, or category-specific volatility rather than a broad disinflation shift. Oil’s roughly 30% July rally directly lifts headline CPI, but core excludes energy, so the relevant channel is indirect pass-through through airfares, freight, input costs, and inflation expectations; those effects usually arrive with a lag and are diluted in the first month. Shelter, medical, insurance, and services inertia still make a sub-0.1% print less likely than usual, while a move above 0.4% would likely require both payback and broad services/goods firmness. Overall I put most mass around 0.2%-0.3%, with thinner tails above 0.4%.
Claude Fable: The 2026 run rate (Jan-May averaging ~0.26%, recent three months ~0.19%) anchors a central expectation near 0.2% MoM. June's -0.02% is an unusual negative print, likely reflecting one-off declines (e.g., used vehicles, airfares, shelter softness); some mean reversion is expected, but it also signals real disinflation momentum, so I center July slightly below the yearly average. The July oil rally (~30%) hits headline directly but reaches core mainly via airfares and transport services with a 1-2 month lag, so it adds only a modest upside skew to July core, maybe 3-5bp. Rounding conventions matter: the print is reported to one decimal, so 'above 0.2' effectively requires an unrounded ~0.25%+. Base rates for monthly dispersion (sd ~0.10-0.12pp around trend) give high confidence the print is positive, roughly even odds of exceeding 0.2%, and low probability of 0.4%+ absent a tariff or shelter shock.
Claude Opus: The six-month core trend (0.30, 0.22, 0.20, 0.38, 0.21, -0.02) averages about 0.215% MoM with high month-to-month noise. June's slightly negative print is almost certainly idiosyncratic , likely a drop in a volatile core component such as airfares, lodging, or used vehicles , and such prints historically revert rather than repeat, so I do not shift the central tendency down much. Shelter, the largest core weight, remains the anchor and has been decelerating only gradually, which keeps a floor near 0.15-0.20%. The 30% July oil rally hits headline immediately but reaches core only indirectly through airfares, delivery costs, and some goods, with a lag of roughly one to three months; only a small fraction shows up in the July print itself. That argues for a distribution centered near 0.22% with a fatter right tail than left. Hence a near-certain positive print, a coin flip above 0.2%, and limited but real probability above 0.3%.
Model estimates generated July 26, 2026, price-blind. These are model estimates, not predictions of fact and not financial or trading advice. Models are frequently wrong; the market price reflects real traders' money. Kalshi is a CFTC-regulated exchange; 18+, availability varies by state.
Related Verdicts
FAQ
Why do the model percentages differ from the Kalshi price?
The models never see the price. When they disagree with the crowd, one side is wrong, and we grade every verdict against real settlements on our scoreboard.
Are model verdicts betting advice?
No. Model verdicts are model estimates, not betting or financial advice. Treat them as one input among many and make your own decisions.
