Fed Rate Decision Odds: What The Models Say
The fed rate decision odds on Kalshi say something remarkable tonight: the crowd's single most likely outcome for the September 2026 FOMC meeting is a rate hike, at 52 cents, with a plain hold trailing at 45 cents. Our panel of six AI models, each scoring the same fetched facts without seeing a single market price, says the crowd is overpaying for that hike. This is the standing OddsShopper page for the Fed decision market: every cycle we run a fresh price-blind panel on the current meeting, and every cycle we come back and grade the last one in public. The September board is below, and so is the honest scorecard from July, where the panel got one contested contract right and one badly wrong.
The Market
| Venue | Kalshi, a CFTC-regulated event-contract exchange (18+; availability varies by state, as of July 2026) |
| The Event | "What will the Fed do at its September 2026 meeting?" sold as five mutually exclusive yes/no contracts: cut more than 25bps, cut 25bps, hold, hike 25bps, hike more than 25bps |
| Settles | On the action the FOMC announces September 16, 2026. Per the written rules, exactly one contract resolves YES; if the meeting were somehow canceled, the hold contract resolves YES |
| Prices Below As Of | The evening of July 29, 2026 (bid-ask midpoints, rounded to the cent), hours after the July decision settled |
One mechanics paragraph and then the numbers, because the numbers are the story. Each contract pays $1.00 if its outcome happens and zero if it does not, so a price in cents reads as a rough implied probability. Rough, not exact: at that snapshot the five raw midpoints summed to about 102.5 cents rather than 100 (the rounded cents in the table below sum a touch higher), which is the board's overround, roughly a 2.5% hold. To de-vig it, divide each raw price by 1.025: the crowd's fair probabilities then read about 51% for the hike and 43% for the hold, and those de-vigged numbers are the right ones to hold up against the model blend below. Kalshi also charges a trading fee that scales with price, about 7% of price times one minus price, rounded up to the next cent, which works out to a cent or two per contract; our full Kalshi fee breakdown has the exact schedule. If prediction markets are new to you, our explainer on how prediction markets work covers the plumbing so this page can stay about the Fed.
The Quick Answer
Kalshi traders make a quarter-point September hike the favorite at 52 cents, with hold at 45 cents and a cut nearly written off. Our six price-blind models read the same meeting the other way: hold is the blended favorite at 57.7%, the hike sits at just 21.7%, and a cut, which the crowd prices at about 4 cents combined, gets a blended 17.9%. The full model-by-model board, the reasoning behind the split, and the graded July scorecard are all below.
The one-look read: The crowd's modal outcome is a hike (52¢). Every one of the six models' modal outcome is a hold (50% to 68%). That is the widest crowd-versus-panel disagreement this page has carried in either of its cycles so far.
The September Board: The Crowd Says Hike, The Models Say Hold
Fresh verdicts, generated the evening of July 29, 2026, hours after the July decision settled. Each model returned one full probability distribution across the five outcomes, price-blind, working from a context card of fetched facts: the official July 29 FOMC statement, the June CPI report, the June jobs report, and same-day meeting coverage.
| Outcome | Kalshi | AI Blend | Claude Fable | Claude Opus | Claude Sonnet | GLM 5.2 | Kimi K3 | DeepSeek V4 |
|---|---|---|---|---|---|---|---|---|
| Cut More Than 25Bps | 1¢ | 2.2% | 2% | 1% | 2% | 2% | 3% | 3% |
| Cut 25Bps | 3¢ | 15.7% | 10% | 12% | 20% | 6% | 16% | 30% |
| Hold | 45¢ | 57.7% | 62% | 68% | 62% | 52% | 50% | 52% |
| Hike 25Bps | 52¢ | 21.7% | 23% | 18% | 14% | 35% | 28% | 12% |
| Hike More Than 25Bps | 3¢ | 2.8% | 3% | 1% | 2% | 5% | 3% | 3% |
These are model estimates, not predictions of fact and not financial advice. Models were scored without seeing market prices. Estimates reflect information available on July 29, 2026; the decision is not until September 16, 2026, and two CPI reports and two jobs reports land in between.
A disclosure on the panel. Our standard Model Verdict panel includes eight seats. Two were unavailable on this machine when these verdicts were generated: the ChatGPT seat (its sign-in was tied up by another machine) and the Gemini seat (the tool was not installed). So this cycle runs on six models, all of which returned complete, coherent distributions.
The row worth staring at is the hike. The blend says 21.7%, but that number papers over the panel's real argument: GLM 5.2 puts the hike at 35%, nearly triple DeepSeek V4's 12%, and those two models read the identical facts in opposite directions. GLM leans into the dissent bloc and the statement's hawkish language; DeepSeek leans into the cooling inflation prints and gives a cut its highest number on the panel, 30%. When the six seats split that wide on both contested outcomes, the honest summary is not "the models say 22%." It is that the panel agrees on the favorite (hold, every seat) and argues about which tail is live.
Why The Models Hold Where The Crowd Hikes
The case for the crowd's hike price starts with what happened hours before these verdicts were generated. On July 29, the Fed held its target range at 3.50 to 3.75 percent on a 9-3 vote, the fifth straight hold, and all three dissenters, Beth Hammack, Neel Kashkari, and Lorie Logan, wanted a quarter-point hike. Coverage flagged it as the most same-direction dissents since September 2016. Behind the dissents sits an oil story: meeting coverage flagged the Middle East conflict as clouding the inflation outlook, energy is the sector the statement itself names in its supply-shock language, and hike odds had been surging alongside oil prices in the week before the meeting. Fed Chair Kevin Warsh, who has deliberately dropped his predecessors' habit of telegraphing the next move, told reporters "I asked for a good family fight and I got one" and said he would not be constrained by outside expectations for September. A visibly hawkish committee with no forward guidance is exactly the setup that gets a hike priced as the favorite.
The models saw all of that in the card and still held, for three reasons that recur across all six analyses.
The dissenters are three votes, not seven. As Claude Fable put it, "a 9-3 vote with three dissents FOR a hike is the strongest same-direction dissent bloc in a decade," but Kimi K3 did the arithmetic the price skips: "converting requires ~4 of the 9 holders to flip within seven weeks, which needs clearly hot data."
The latest data leans the wrong way for a hike. June core CPI cooled to 2.6% year over year from May's 2.9%, with headline CPI down 0.4% on the month, and June payrolls grew just 57,000 against roughly 115,000 expected, with April and May revised down a combined 74,000. Claude Sonnet's read: "June's +57k miss and 74k of downward revisions were already known at the July meeting yet produced zero cut votes, so further, clearer labor deterioration is needed to tip the balance." That sentence is why Sonnet has the panel's lowest hike number among the Claude seats and still refuses to make a cut the favorite.
Committees that just held five straight times tend to hold. Claude Opus anchored on the base rate: the Fed rarely moves without telegraphing, Warsh gave no September signal, and so "hike risk clearly outweigh[s] cut risk barring sharp labor deterioration," with hold at 68%, the panel's high.
And the swing factors the models circled all land in the same pre-meeting stretch.
The swing releases: August CPI, due around September 10-11, days before the decision, and the early-September jobs report. GLM 5.2 was explicit that "a core reacceleration would push hike probability above 50%"; Claude Opus instead named the August jobs report, where "a sharply negative print flips the balance toward a cut." Both are worth remembering when this page gets rebuilt after September 16.
The dissent bloc is the crowd's whole case; the models' answer is that dissents are a warning light, not a majority.
What One Meeting Did To The Models' Read
Here is the part that makes a standing page worth keeping. We also ran this September panel before the July decision, and that earlier run, published on our September preview page, had a completely different shape: a 25bps cut was the blended favorite at 48%, hold sat at 30%, and the hike was a 7% afterthought. One meeting later, with the hold, the three hawkish dissents, and the "family fight" presser in the card, the panel's 25bps-cut estimate collapsed from 48% to 15.7% and its 25bps-hike estimate tripled from 7% to 21.7%. (Two caveats for honesty: the earlier run carried eight seats to this run's six, and both runs are equal-weight blends, so seat composition moves the number a little. The direction of the swing is not subtle either way.) The models are not stubborn, and that is the point of re-running them every cycle instead of letting a stale verdict sit on a permanent URL.
Last Fed Decision: How The Models Did
The July 2026 contracts settled on July 29: the Fed held, so "Fed maintains rate" resolved YES and everything else expired worthless. Our panel scored that board on July 24, five days out. Here is the full grade, market price as of the run date next to the panel's blend.
| Outcome | Kalshi (Jul 24) | Panel blend | Result | Grade |
|---|---|---|---|---|
| Cut More Than 25Bps | 1¢ | 2% | NO | Both right |
| Cut 25Bps | 1¢ | 17% | NO | Market right, panel wrong |
| Hold | 75¢ | 66% | YES | Both right; market more confident |
| Hike 25Bps | 26¢ | 12% | NO | Panel right, market overpriced it |
| Hike More Than 25Bps | 1¢ | 2% | NO | Both right |
Both the crowd and the panel made hold the clear favorite, and hold happened; the market's 75 cents was closer to the truth than the panel's 66%. The two contested rungs split. On the hike, the panel's fade was vindicated: it said 12% while the crowd held the contract at 26 cents five days out (it still traded at 18 cents just before close), and the contract died. On the cut, the panel embarrassed itself: a blended 17% for an outcome the market correctly priced at 1 cent. Several models had talked themselves into the soft jobs data forcing the Fed's hand early. It did not, and the crowd never believed it would.
That cut miss is the one I keep coming back to, because it is the same disagreement the panel is having again right now: DeepSeek's 30% cut estimate for September is the July mistake's shape, one meeting later. Either the labor data finally deteriorates enough to make the models' cut lean look early rather than wrong, or the crowd buries it a second time. That is precisely what this page's next edition will grade.
The Track Record
One row per completed cycle, market favorite versus panel favorite, and who was right. Full postmortems stay on this page for the most recent cycles; older ones collapse into this table so the current verdict never gets buried.
| Cycle | Market favorite (at run) | Panel favorite (blend) | Result | Read |
|---|---|---|---|---|
| July 2026 | Hold, 75¢ | Hold, 66% | Hold | Both hit; panel better on the hike rung, market far better on the cut rung |
One cycle is not a sample. The table exists so that by the time it is six rows deep, you will not have to take our word for how the panel performs; you can count.
How The Fed Decision Market Works
The evergreen part, for readers arriving from a search like "will the Fed cut rates" rather than from prediction-market land. Kalshi's Fed decision market is a set of event contracts on the FOMC's announced action, one contract per outcome, re-listed for every scheduled meeting; exactly one contract in the set resolves YES, the written settlement rules govern every edge case (Kalshi's rules even specify that a canceled meeting resolves the hold contract YES), and prices move around the clock as data lands. Contract mechanics in full are their own guide, how Kalshi's economic markets work, so here it is enough to say what the structure buys you. That standing structure is what makes the board a running, money-backed projection of the next FOMC meeting, why "FOMC meeting odds" and Fed funds futures usually tell a similar story, and why a 30-point gap between the crowd and an independent panel, like this cycle's hike row, is worth an article at all. It is also what makes the market re-gradable: this page has already watched a single meeting move its own 25bps-cut estimate by 32 points, which no static explainer of the Fed can show you.
A Worked Example From The July Cycle
A trader who bought one "Fed maintains rate" contract at its July 24 price of 75 cents was paid $1.00 at settlement on July 29, a 25-cent gross profit, less Kalshi's trading fee of two cents at that price (the formula rounds up to the next cent). Had the Fed hiked instead, the same contract would have expired worthless and the full 75 cents would have been lost. There is no partial credit in event contracts, which is exactly why the probabilities on this page are about handicapping the announcement itself, not the economy in general. For the broader comparison of venues that list markets like this one, see our ranking of prediction market platforms; for a stranger cousin of this market, the panel has also scored whether the Fed holds an emergency meeting before 2027.
One honesty note about this site: OddsShopper's live odds screen and odds comparison tools cover sports markets, not Fed contracts, so this page pitches no tool. It runs the panel, shows the board, and keeps score. Kalshi operates as a CFTC-regulated exchange, open to users 18 and older, with availability that varies by state as of July 2026. Nothing on this page is advice about whether or how to trade any of it.
FAQ
Will The Fed Cut Rates At The September 2026 Meeting?
The market says almost certainly not: Kalshi prices the two cut contracts at about 4 cents combined. Our six-model panel is warmer on the idea at a blended 17.9%, mostly on the stalling payroll numbers, though on the blend a cut still trails both a hold and a hike. Two seats do rate a cut ahead of a hike once you combine both cut rungs and both hike rungs: Claude Sonnet (22% total cut versus 16% total hike) and DeepSeek (33% versus 15%).
What Are The Odds Of A Fed Rate Hike In September 2026?
As of the evening of July 29, 2026, the hike-25bps contract was priced at about 52 cents on Kalshi, the top price on the board. The price-blind panel blends that same outcome at 21.7%, the widest crowd-versus-models gap on the board.
Why Does The Market Think The Fed Will Hike?
Three FOMC members, Beth Hammack, Neel Kashkari, and Lorie Logan, dissented at the July meeting in favor of a quarter-point hike, the most same-direction dissents since September 2016, and the statement stressed that inflation remains elevated relative to the 2 percent goal. Traders read that as a committee drifting hikeward under a chair who refuses to signal.
How Did The Models Do On The Last Fed Decision?
Both panel and market correctly favored a July hold. The panel beat the crowd on the hike contract (12% against a 26-cent price, resolved NO) and lost badly on the cut contract (17% against a 1-cent price, also NO). The graded table above has the full rung-by-rung read, and this page re-grades itself every cycle.
Are These Model Estimates Financial Advice?
No. They are model estimates, not predictions of fact and not financial advice, generated price-blind from fetched public data. Treat them as one input among many and make your own decisions.
The Bottom Line
Warsh asked for a good family fight and got one, and the market promptly priced the dissenters as the next majority. Six models, reading the same statement and the same data with no prices in view, all landed on the quieter answer: five holds tend to become six, three votes are not seven, and the inflation prints are cooling rather than forcing anyone's hand. Somebody is wrong by nearly 30 points of probability on the hike contract (a de-vigged 51% against the panel's 21.7%), which is as clean a test of crowd versus machines as this series has produced. The last data worth waiting for lands in the week and a half before the decision: the early-September jobs report and then the August CPI print around September 10-11, the two releases the models named as capable of flipping their answers. On September 16 the Fed will settle the argument, this page will grade both sides in the table above, and the next cycle's board will take this one's place. If sports is more your speed while the Fed keeps everyone waiting, our experts post free picks every day.
Kalshi is a CFTC-regulated event-contract exchange (18+, availability varies by state). This article is informational and does not constitute financial, investment, or trading advice.
