Core CPI Odds: Did July Inflation Cool? The Print Vs. The Market

THE PRINT IS IN (updated ~8:45 AM ET, Aug 13): July core CPI (all items less food and energy, BLS series CUUR0000SA0L1E) came in at 2.48% year over year — index 337.133 vs 328.98 a year ago — and +0.07% month over month. Headline CPI printed 3.36% YoY. Everything below was written against the pre-print market; the graded read on how the market priced it follows in tonight's update once Kalshi's settlement posts.

Every month, one government number decides a full ladder of Kalshi contracts, and almost all of the action comes down to a single rounding line. The Bureau of Labor Statistics reports core CPI to one decimal place, so the difference between a reported 0.2% and a reported 0.3% is a few thousandths of a percentage point on the underlying monthly change, and it is worth the entire face value of the "above 0.2%" contract. Two and a half weeks ago, our price-blind AI panel and the Kalshi crowd took opposite sides of exactly that line on the July print. The crowd won, and the graded table further down this page shows precisely how.

This is the standing core CPI page. Each cycle it carries the panel's fresh read on the next print, the scored postmortem of the last one, and a plain-English guide to how the market itself works. The current cycle is August 2026 core CPI, which settles on the September 11 release.

The Quick Answer

The seven-model panel puts a 76% chance on the August core CPI print coming in above 0.1% and a 33% chance on it clearing 0.2%, with almost nothing above 0.4%. That is a distribution centered on a reported 0.2%, right where the Cleveland Fed's +0.20% nowcast sits: mid-bucket, five unrounded basis points below the line that would round the print up to 0.3%. The full board, every seat's number, and the July postmortem where the crowd beat the panel on this exact rung are all below.

Free: The Weekly PM Market Brief — the panel's graded record, the week's biggest market-vs-model gaps, and what settles next. When this board is graded on September 11, the scored result lands there first. One email, Sundays.

A timing note before the board: the July CPI report landed the morning of August 12, and by Wednesday evening the August ladder still had not printed a single trade since the release. Prices below are the latest trades as of the evening of August 12, 2026, ET, with the live bid and ask noted where a quote looks stale. This page is rebuilt as the story moves, and the "as of" stamps carry the evidence.

The Board: August Core CPI, Rung By Rung

August Print Above	Kalshi last trade	AI blend (7 models)
0.0%	99¢	95%
0.1%	85¢	76%
0.2%	36¢	33%
0.3%	20¢	8%
0.4%	10¢	1.5%
0.5%	1¢	0.3%

(The event lists rungs up to 1.0%. The five rungs above 0.5% are omitted here; at our evening check none carried a live bid, and the 0.5% rung itself last traded at 1¢ against a 0-bid, 15¢-ask book.)

The row that matters is 0.2%. The panel's blend implies the single most likely reported value is exactly 0.2%, with about a 43% chance the print lands right on it. The market column needs an honest caveat, though: the 36¢ last trade on that rung predates the July release, and the live book at our evening check was a 23¢ bid against a 69¢ ask. The panel's 33% sits inside that spread, but there is no post-print crowd view to compare it to yet; nobody has traded this board since the number dropped. In July the same rung carried a 40-point disagreement; whether the two sides have actually converged is something only the next real trades can show.

The tail rows are the same story, only starker. The 0.3% rung last traded at 20¢, but the live book at our evening check showed a 3¢ bid against a 42¢ ask, and the 0.4% rung's 10¢ last trade sits on a 3¢-bid, 30¢-ask book. The panel prices those rungs at 8% and 1.5%. The two bottom rungs are no better: the 99¢ last trade sits on a 60¢-bid, 100¢-ask book and the 85¢ on 50/89, so treat both as marks, not prices. Until real volume returns after the July release, the last trades on this ladder are old wood, not the crowd's current opinion.

New to how these contracts work? The companion guide to Kalshi's Fed, CPI and jobs markets covers the plumbing in full; the short version is one section below.

The Market At A Glance

Venue	Kalshi, a CFTC-regulated event-contract exchange (18+; availability varies by state, as of August 2026)
The Event	"CPI core in August" (Kalshi event KXCPICORE-26AUG), yes/no contracts at thresholds from 0.0% to 1.0%
The Contract	Each market resolves YES if seasonally adjusted core CPI (all items less food and energy) for August 2026 rose by more than the stated threshold, per the contract terms. The settlement value is the single-decimal number the Bureau of Labor Statistics reports
Settles	Friday, September 11, 2026, on the 8:30 a.m. ET release (the market closes at 8:25 a.m. ET). A government-shutdown data delay extends expiration under the contract rules
Prices As Of	August 12, 2026, evening ET. The board is thin and had not traded since the July release at our evening check, so the live bid and ask are quoted wherever a last trade looks stale

Every Seat's Number

Seven of the panel's eight seats ran this board; the Gemini seat has been down since an August 9 outage and sat this cycle out. That roster note matters for the July comparison below: July's blend came from all eight seats, so part of any move between the two cycles is a changed roster, not just changed minds. Each model priced the whole ladder in a single pass from one distribution, price-blind, from a fetched card of BLS index data, the July release detail, the Cleveland Fed nowcast, and the contract's rounding rules. These are the numbers after a revision round in which each model read the others' anonymized reasoning.

August Print Above	Kalshi	AI blend	ChatGPT (GPT-5.5)	Claude Fable	Claude Opus	Claude Sonnet	GLM 5.2	Kimi K3	DeepSeek V4
0.0%	99¢	95%	96%	95%	95%	97.5%	96%	92%	94%
0.1%	85¢	76%	77%	74%	74%	80%	82%	70%	78%
0.2%	36¢	33%	32%	33%	35%	32%	36%	34%	31%
0.3%	20¢	8%	8%	8%	9%	6%	10%	8.5%	7%
0.4%	10¢	1.5%	1.3%	1.5%	1.8%	1%	2.5%	1.3%	0.8%
0.5%	1¢	0.3%	0.25%	0.3%	0.4%	0.15%	0.5%	0.1%	0.1%

Every seat on this panel is graded against real market settlements — records to date: GPT 83% on 5,138 graded calls · Claude Fable 82% on 460 graded calls · Claude Opus 84% on 524 graded calls · Claude Sonnet 83% on 508 graded calls · GLM 81% on 2,318 graded calls · Kimi 83% on 2,296 graded calls · DeepSeek 80% on 2,343 graded calls. Each percentage is directional accuracy: the share of settled yes/no contracts where the seat's stated probability sat on the side that actually settled. Recomputed daily; the full scoreboard is public.

Model estimates generated August 12, 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.

Every number in this piece gets graded in public once the market settles, on the same scoreboard linked above.

More live boards from the same panel: the Fed's September decision (hold last traded 67¢) · the August jobs ladder (above 50,000 jobs, 57¢ bid) · the headline CPI board.

Why The Panel Centers On 0.2%

The case is base rates first. The 2026 core CPI prints, unrounded, run +0.295, +0.216, +0.196, +0.376, +0.208, minus 0.017, and +0.215. Four of the seven sit in a tight band from +0.196% to +0.216% (February, March, May and July), and the three escapes went in three directions: January's +0.295% and April's +0.376% ran hot, and June went flat. The two hot ones would have rounded to 0.3% or higher — a 29% base rate for the hot side, landing within a few points of the panel's 33% from a completely different direction than the nowcast. The strongest counter runs through the same list: strip June's outlier and the remaining six average +0.251%, sitting exactly on the rounding line. A process like that, centered near +0.21% with modest dispersion, makes a reported 0.2% the single most likely August outcome and leaves real but limited weight on 0.3%.

Cleveland Fed nowcasts +0.20% and 2026 prints cluster tightly at 0.20-0.22% unrounded, with cool shelter capping upside and rounding making a reported 0.3% a one-in-three tail. — Claude Fable

The second leg is shelter, the heaviest component in core. The reported figure held at +0.1% in both June and July, and unrounded it ticked from +0.118% to +0.139%. Several seats called shelter the dominant anchor: as long as it rounds to a tenth a month, the swing factors, used cars at +0.4% and airfares at +2.2% in July, have to do a lot of work to drag the whole index to a reported 0.3%. One honest caveat on reading that tick as a trend: owners' equivalent rent ran +0.297% in May, +0.240% in June and +0.260% in July, so the June-to-July uptick sits inside normal month-to-month noise. Shelter would need roughly one more move of July's size to round up to +0.2% — worth watching, not yet a trend.

The third leg is the nowcast. The Cleveland Fed nowcast on the panel's fetched card, the August 6 vintage, put August core at +0.20%, and its July estimate of +0.21% landed within half a basis point of the actual +0.215%. That track record earned it the anchor role in most seats' reasoning.

Held because no peer surfaced new evidence beyond the Cleveland Fed nowcast and component detail I already weighed; the knife-edge nowcast at +0.20% aligns with my central 0.2% mass and a ~31% tail probability of rounding up. — DeepSeek V4

What keeps the 0.2% rung at 33% rather than lower is rounding mechanics. A center of +0.20% means the print only needs to drift about five unrounded basis points hot to round up to 0.3%, and this series has produced bigger upside surprises than that twice this year.

The panel converged on my own read — center ~+0.21% unrounded with the rounding line at +0.25% — and no peer surfaced evidence I had not already weighed, so I hold with a hair more right tail for lagged energy pass-through. — Claude Opus

Where The Panel Changed Its Mind

The revision round moved the edges, not the center. GLM came in highest on the upper tail and trimmed it, cutting its 0.3% rung from 12% to 10% and its 0.4% rung from 4% to 2.5%: "I trimmed my right tail after peers highlighted that volatile swing factors (airfares +2.2%, used cars +0.4%, medical [care services] +0.6%) tend to mean-revert rather than repeat, and that the August nowcast vintage is early-in-cycle with wider error bands than the accurate July read." Claude Sonnet moved the other way, thickening its 0.3% number from 3.5% to 6% — its center "stays near 0.20-0.21% unrounded on cooling shelter, but the right tail needed thickening." Kimi split the difference, nudging its center from +0.21% to +0.205% unrounded and trimming its money rung from 36% to 34% after peers argued the two straight +0.1% shelter prints and the +0.20% nowcast pin the mean closer to 0.20.

The most instructive non-move came from Claude Fable, which held every rung and said why: "my center (+0.21% unrounded, sd ~0.10pp) already reflects the peer arguments, and the panel's spread brackets my numbers symmetrically." All seven seats finished within a five-point band on the money rung, 31% to 36%, which is about as close to consensus as this panel gets.

What Would Change The Panel's Mind

Three checkable things between now and September 11, each with a direction:

• The Cleveland Fed's Daily Nowcast Vintage. It updates daily through the reference month. A drift from +0.20% toward +0.25% or higher would push the 0.2% rung's fair value up sharply, because that is exactly the rounding line; a drift toward +0.15% would collapse it.
• The August Jobs Report, September 4. Average hourly earnings at +0.3% or hotter would firm the services-inflation story and nudge the upper rungs; +0.1% or cooler leans the other way.
• Shelter's Run Rate. Two straight reported months at +0.1% is the panel's anchor. Unrounded, July ran +0.139%; one more move of that size rounds shelter to +0.2%, the whole distribution shifts, and this page gets re-scored.

One input that will not move this board: oil. Crude is excluded from core and reaches airfares and freight with a one-to-three-month lag, so even a sharp re-spike matters more for September's ladder than this one. Worth knowing mainly so an oil headline doesn't tempt you into repricing rungs it cannot touch.

Last Core CPI Print: How The Models Did

That cycle settled on the morning of August 12: core CPI printed +0.2% on the month (+0.215% unrounded), so every contract from "above 0.2%" upward settled NO. Here is the full graded board, with the market prices and the eight-seat panel blend as they stood when the verdicts were generated on July 26.

July Print Above	Market then	Panel blend then	Result	Who was right
0.0%	96¢	91%	YES	Both
0.1%	77¢	77%	YES	Both, to the cent
0.2%	11¢	51%	NO	The crowd
0.3%	22¢	24%	NO	Both
0.4%	2¢	8%	NO	Both
0.5%	6¢	2%	NO	Both

One tell worth naming in our own table first: the late-July market column is not internally consistent — above-0.3% (22¢) marked higher than above-0.2% (11¢), and above-0.5% (6¢) above above-0.4% (2¢), which is impossible for cumulative rungs and means the tail marks were stale, thin last trades rather than one coherent crowd view. The 11¢ on the money rung was the traded price, and that is the rung the grade stands on.

The honest read: the panel got the shape right and the bet wrong. Its distribution said the July print centers between 0.2% and 0.3%, and the actual unrounded value, +0.215%, landed inside that call. But the tradeable question was whether the print would clear the rounding line at roughly +0.25% unrounded, and the panel put it at a coin flip while the crowd priced it at 11 cents. The print missed the line by about three and a half unrounded basis points. Near miss or not, a graded miss is a miss, and it goes on the scoreboard like every other call.

That July miss is also why this cycle's board looks different. Handed the same knife-edge one month later, the panel — now seven seats rather than July's eight — prices "above 0.2%" at 33% rather than 51%, not because it read the market (it never sees a price) but because July's settled print gave it one more data point saying the center of this series is 0.20%, not 0.25%. Whether that is a lesson learned or a lesson over-learned gets graded on September 11.

How The Core CPI Market Works

Kalshi lists one event per month in the KXCPICORE series, with yes/no contracts at each threshold. Everything settles on a single number: the seasonally adjusted month-over-month change in CPI for All Urban Consumers, All Items Less Food and Energy, as the BLS publishes it, to one decimal. "Above 0.2%" means a reported 0.3% or higher. A print of exactly 0.2% settles NO, which is precisely how the July contract died.

The rounding convention is the whole game: the true dividing line for "above 0.2%" sits near +0.25% unrounded, so a nowcast centered at +0.20% — a reported value, not a dividing line — leaves the next rung's fate to the width of the distribution around it. It is also the one place the panel gives you a number the book doesn't: if the print has a 33% chance of clearing +0.25% unrounded, fair value on the 0.2% rung is 33¢, against a live book quoted 23¢ to 69¢. Where the fair number sits inside a spread that wide is exactly what a reader of this page knows that a reader of the raw board does not.

Two quirks worth knowing. First, the contract terms carry a shutdown clause: if a federal shutdown delays the data, expiration extends until the release. The risk isn't theoretical: BLS never published an October 2025 CPI report during that fall's shutdown, leaving a permanent hole in the series; the October 2025 line is still blank in the BLS database today. Second, core excludes food and energy, so an oil move never hits this board directly; it seeps in through airfares and freight over one to three months. Contracts trade on Kalshi's book with real spreads, and on quiet boards like this one the last trade can be days old, so check the live bid and ask before treating any price on this page as current.

A Worked Example: How July Settled

July's core CPI index came in at 336.789 against June's 336.065, a rise of +0.215%. BLS reported that as +0.2%. Walk the ladder with that one number: "above 0.0%" and "above 0.1%" settled YES and paid holders $1 per contract; "above 0.2%" and everything higher settled NO and paid the no side. A trader who bought the 0.2% rung's YES at the 11¢ it was trading at in late July lost the stake; a trader who sold it, or bought NO, collected. One decimal place decided the entire ladder, with no judgment calls anywhere.

Settlement Timeline

Date	What happens
Daily	Cleveland Fed inflation nowcast updates; the panel's anchor number moves with it
September 4, 2026	August jobs report, the last major inflation-adjacent release before settlement
September 11, 2026	BLS publishes August CPI at 8:30 a.m. ET; the market closes at 8:25 a.m., expires by about 10:00 a.m. ET, and the board gets graded
September 15-16, 2026	FOMC meeting, four days after settlement, with hold last traded at 67¢ on Kalshi

This page is rebuilt each cycle at this URL, and re-scored when the story moves between releases.

FAQ

When is the next CPI report?

The next release is Friday, September 11, 2026, at 8:30 a.m. ET, covering August data. CPI comes out around the middle of each month, one month in arrears, on the Bureau of Labor Statistics schedule.

What did the last CPI report show?

The July 2026 report, released August 12, showed headline CPI up 0.1% on the month and 3.4% year over year, with core CPI up 0.2% on the month and 2.5% year over year.

Hottest Prediction Markets Right Now

• 2028 Democratic presidential nominee · $177M traded
• 2027 Pro Football Champion · $59M traded
• 2028 U.S. Presidential Election winner? · $59M traded
• 2028 Republican presidential nominee · $56M traded
• Pro Baseball Champion · $52M traded
• More tech layoffs in 2026 than in 2025? · $31M traded

Every market above links to our full AI model verdict; browse them all on the OddsShopper prediction markets hub, and see how every settled call actually scored on the full graded scoreboard.

The Bottom Line

The August core CPI board is a study in patience. The ladder has not traded since the July print, so the only fresh number on it tonight is the panel's: 33% that August clears the 0.2% line, centered on the same +0.20% nowcast the crowd will eventually price against. The July postmortem above is the reason to trust the process rather than any single call: the panel published a coin flip, the market said 11 cents, the market was right, and the grade is public. If you want the graded result the morning this board settles, the Weekly PM Market Brief carries it first, and our free expert picks run daily in the meantime. On September 11 at 8:30 a.m. ET, one decimal place settles everything above, and this page will say so either way.