2028 Democratic Nominee Odds: The AI Model Verdict Vs. The Kalshi Market

The 2028 Democratic nomination market is the biggest political board on Kalshi: 46 named candidates and roughly 168 million contracts traded, more than the Republican nomination board and the 2027 pro football championship market combined. We put all 46 names in front of a panel of six AI models, handed them a fresh file of real data, hid every market price from them, and asked one question: who actually wins this thing? The models agree with the crowd about the favorite almost to the cent. Then they disagree with it about nearly everyone behind him, in both directions, and the reasons why are the fun part.

The Quick Answer

Six AI models, working price-blind from fetched polling, office, and settlement data, make this race a dead heat at the top: Gavin Newsom at a 16.9% blend against his 16.5-cent Kalshi price, and Kamala Harris at 16.3% against just 9.4 cents. Behind them the disagreements get loud. The panel holds Pete Buttigieg at 13.8%, more than double his 5.5-cent price, powered by the early-state polling. It fades Alexandria Ocasio-Cortez to 9.7% against her 13.5-cent price. And it prices Jon Ossoff at 2.5% against 14.5 cents, still the widest single disagreement on the board. The full 46-name board, every seat's number, and the dark horses the panel refuses to dismiss are all below.

Price Check: August 3, 2026

The panel's verdicts and blend are frozen at their generation date, August 3, 2026; prices quoted in the prose are verdict-day prints. This block and The Board refresh from Kalshi's live order book.

Free: The Weekly PM Market Brief — the AI panel's graded record, the week's biggest market-vs-model gaps, and what's spiking next. One email, Sundays. One email, Sundays — the signup box is at the bottom of this page, or just keep reading: the record it promises is on this one.

• Market top five right now: Gavin Newsom 16.5c, Jon Ossoff 14.5c, Alexandria Ocasio-Cortez 13.5c, Kamala Harris 9.4c, Mark Kelly 6.2c
• Biggest moves since verdict day: no market has moved 0.5c or more since August 3, 2026
• Widest live gaps vs the frozen blend: market above the panel — Jon Ossoff (14.5c vs 2.5%); market below the panel — Pete Buttigieg (5.5c vs 13.8%)
• Board totals: about 167.9 million contracts traded, 114 million open

The Board

Panel blend is the equal-weight average of each seat's final distribution after the revision round, frozen August 3, 2026. Kalshi prices are quote midpoints from the live book as of August 3, 2026; Δ is the move since the verdict-day print. Office descriptions were fetched at verdict time from the congress-legislators dataset and Wikipedia.

Candidate	Current Office	Kalshi (August 3, 2026)	Δ since August 3, 2026	AI Blend
Gavin Newsom	Governor of California (since 2019)	16.5c	—	16.9%
Kamala Harris	Former Vice President (2021-25)	9.4c	—	16.3%
Pete Buttigieg	Former Transportation Secretary (2021-25)	5.5c	—	13.8%
Alexandria Ocasio-Cortez	U.S. Representative, NY	13.5c	—	9.7%
Josh Shapiro	Governor of Pennsylvania (since 2023)	4.5c	—	6.6%
The Field (anyone unlisted)	–	not offered	—	6.1%
Mark Kelly	U.S. Senator, AZ	6.2c	—	4.2%
Andy Beshear	Governor of Kentucky (since 2019)	3.8c	—	3.6%
Gretchen Whitmer	Governor of Michigan (since 2019)	1.0c	—	2.9%
Jon Ossoff	U.S. Senator, GA (term ends Jan 2027)	14.5c	—	2.5%
Wes Moore	Governor of Maryland (since 2023)	2.5c	—	2.1%
Cory Booker	U.S. Senator, NJ	0.7c	—	2.0%
J.B. Pritzker	Governor of Illinois (since 2019)	2.6c	—	1.7%
Bernie Sanders	U.S. Senator, VT	0.7c	—	1.1%
Tim Walz	Governor of Minnesota (since 2019)	0.1c	—	1.1%
Raphael Warnock	U.S. Senator, GA	0.5c	—	0.9%
Elissa Slotkin	U.S. Senator, MI	0.5c	—	0.8%
Ruben Gallego	U.S. Senator, AZ	0.5c	—	0.8%
Chris Murphy	U.S. Senator, CT	0.7c	—	0.7%
Ro Khanna	U.S. Representative, CA	2.4c	—	0.7%
Amy Klobuchar	U.S. Senator, MN	0.1c	—	0.6%
Jared Polis	Governor of Colorado (since 2019)	0.1c	—	0.5%
Michelle Obama	Former First Lady (2009-17)	1.2c	—	0.5%
Zohran Mamdani	Mayor of New York City (since 2026)	0.2c	—	0.5%
Rahm Emanuel	Former Chicago mayor; ambassador to Japan (2022-25)	5.1c	—	0.5%
Elizabeth Warren	U.S. Senator, MA	0.1c	—	0.4%
John Fetterman	U.S. Senator, PA	0.1c	—	0.4%
Roy Cooper	Former Governor of North Carolina (2017-25)	0.2c	—	0.4%
James Talarico	Texas state representative; 2026 U.S. Senate nominee (TX)	1.6c	—	0.3%
Jasmine Crockett	U.S. Representative, TX	0.1c	—	0.2%
Gina Raimondo	Former Commerce Secretary (2021-25)	0.1c	—	0.2%
Phil Murphy	Former Governor of New Jersey (2018-26)	0.1c	—	0.2%
Mitch Landrieu	Former New Orleans mayor (2010-18)	0.5c	—	0.2%
Beto O'Rourke	Former U.S. Representative, TX (2013-19)	0.1c	—	0.1%
Abdul El-Sayed	Former Wayne County health director; epidemiologist	0.5c	—	0.1%
Jon Stewart	Comedian and TV host	1.2c	—	0.1%
Hillary Clinton	Former Secretary of State (2009-13)	0.1c	—	0.1%
Mark Cuban	Businessman	1.0c	—	0.1%
Andrew Yang	Forward Party founder	0.2c	—	0.1%
Graham Platner	Oyster farmer; ended 2026 ME Senate campaign in July	0.1c	—	0.1%
Stephen A. Smith	ESPN television personality	0.5c	—	<0.1%
Jamie Dimon	Chairman and CEO, JPMorgan Chase	0.2c	—	<0.1%
Dwayne Johnson	Actor and professional wrestler	0.1c	—	<0.1%
Liz Cheney	Former U.S. Representative, WY (2017-23)	0.1c	—	<0.1%
Hunter Biden	Businessman; son of former President Joe Biden	0.7c	—	<0.1%
LeBron James	NBA player, Philadelphia 76ers	0.1c	—	<0.1%
Barack Obama	Former President (2009-17)	0.2c	—	<0.1%

The Market

Venue	Kalshi — a CFTC-regulated event-contract exchange (18+; availability varies by state, as of August 2026)
The Contract	One market per candidate, 46 listed: "If [candidate] wins and accepts the nomination for the Presidency for the Democratic party in 2028, then the market resolves to Yes." Kalshi's contract terms define the nominee as the first person to be nominated by the party and to accept; a nominee who later withdrew would still settle Yes, and post-expiration revisions do not count.
The Clock	Markets close early once the party nominates someone and they accept, at the 2028 Democratic convention; the listed expiration is November 7, 2028
Depth	About 167.9 million contracts traded across the board, 114 million open, as of August 3, 2026. Quotes run in 0.1-cent ticks below 10 cents, so even sub-1% names carry two-sided books, though the thinnest few trade on spreads nearly as wide as their price
Timestamps	Verdicts generated August 3, 2026 (both rounds); prices as of August 3, 2026, from the live order book

One structural note before the stories: the 46 listed names are Kalshi's listing choices, not a ballot, and there is no catch-all contract. If someone not on the board is nominated, every listed market settles No. The panel was therefore required to price a 47th bucket, the unlisted field, and gave it real weight: 6.1%.

How Six AI Models Priced This Race

The panel is six models: Claude Fable, Claude Opus, Claude Sonnet, GLM 5.2, Kimi K3, and DeepSeek V4. Each one received the same file of freshly fetched facts, what we call the data card: the full 46-contract board with its binding settlement language, every candidate's current office pulled live from government and encyclopedia sources, and every public national poll of the 2028 Democratic primary from the last 90 days, drawn from two independent polling databases and cross-checked against each other, plus every statewide primary poll published in the same window. Each poll came annotated with its pollster's graded accuracy record, so a model could tell a sharpshooter from a coin-flipper.

What no model saw: a single market price. That is deliberate. A forecaster who knows the market's answer tends to drift toward it, so we hide the answer and make each seat price the whole field from scratch, as a set of probabilities that must sum to 100. The headline number for each candidate is the blend, a simple equal-weight average of the six seats. When the blend and the market disagree, that gap is the story, and sometimes the market is the one that's right. At the White House Correspondents' Dinner in July, the Kalshi crowd narrowly out-forecast our best seat across 30 settled markets, and we published the full scorecard anyway. Every number in this piece gets graded in public once the market settles, on the full graded scoreboard.

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.

The Polling Argument That Splits The Board

Here is the puzzle every seat had to solve. Kamala Harris leads 25 of the 29 national primary polls in the card, and not narrowly: 29.7% with Big Data Poll in late July, 31% with J.L. Partners, 37% with Harvard Harris. A naive read says she is the runaway favorite.

The best-graded pollsters complicate that read badly. The card grades every pollster by its historical average miss, and the two sharpest records in the file split on her. TIPP Insights (2.15-point average miss) has Harris first, at 31. But AtlasIntel (2.21) inverts the entire board: Ocasio-Cortez 26, Buttigieg 22, Newsom 21, Harris 13. Emerson College's July poll has her at 8, in fifth.

Then come the early states, and the bottom falls out. The University of New Hampshire's July poll has Harris at 5% among Granite State Democrats, behind Ocasio-Cortez at 22 and Buttigieg at 21. Saint Anselm's June poll of the same state: Harris 4, Buttigieg 29. Primary voters who are actually paying attention this early are not choosing the name they know best, and New Hampshire voters see more candidates up close than anyone. The one counterweight is Michigan, where a Detroit News/WDIV poll has Harris outright first at 25. The panel spent both of its rounds arguing about which of these signals to believe, which is exactly what a 16.3-versus-9.5 disagreement with the market looks like from the inside.

"Early polls measure name ID, not commitments; Harris's wild cross-pollster variance and New Hampshire collapse signal soft support…" — Kimi K3

Kamala Harris: Nine Cents For A 25-Poll Leader

The market's Harris price is the board's quiet shock. She leads nearly every national poll, and you can buy her Yes for 9.4 cents, which is the crowd saying she has less than a one-in-ten chance. The panel blends her at 16.3%, a 6.9-point disagreement with the money that only one number on the board exceeds.

The honest version of the bull case is not "the polls say she wins." It is that even heavily discounted name-recognition support is still support, and the Michigan poll shows her lead surviving in at least one early, engaged electorate. The bear case is everything in the section above. What makes her number the panel's most fragile is how far apart the seats still sit after debating it: Opus at 12, GLM at 20.5. No other top-tier candidate divides them like that.

Pete Buttigieg: The Panel's Strongest Disagreement

Buttigieg at 5.5 cents is that one wider disagreement: the panel's most confident quarrel with the market anywhere on the board. Every seat prices him in double digits, 12 at the floor, for a 13.8% blend at roughly two and a half times his market price. The case writes itself from the card: first place in Emerson's July national poll at 19, second in AtlasIntel, and the New Hampshire prints, 21 and a field-best 29, that no other candidate approaches. The market is paying single digits for the field's best early-state polling, and the panel's estimate says that number is several sizes too small.

Jon Ossoff: The 14-Cent Question

The Georgia senator is the market's co-second favorite at 14.5 cents, and the panel's answer is 2.5%. That is the widest gap on the board, and it deserves the most careful explanation, because the market has receipts here.

For most of the summer, Ossoff barely registered in primary polling, typically 1 to 3 percent in the polls that listed him at all. Then in July, Emerson added him to its ballot test for the first time and he entered at 13%, tied for third. This page's previously published board, generated July 30, blended him at 1.3%; the crowd was at 14.5 cents then too, and the crowd, it turned out, was already trading on the shape of exactly that kind of news. This is what "the market knows things" looks like in practice, and it is why we grade ourselves in public.

So why does the panel still fade him this hard? Because one poll is one poll. Most of the other national surveys that list him stay at 1 to 3 percent, and the newest, Big Data in late July, prints him at 6.2, still under half the market's number. His 14.5 cents now sits ahead of everyone except Newsom on the board, and his nearest catalyst is a hurdle, not a headline: his Senate term ends in January 2027, which means he first has to win reelection in Georgia in November 2026 before any national campaign story can begin. Several seats treat that race as the whole trade. Fable's dark-horse note captures the both-ways read: "Emerson's rated poll printed him at 13 nationally, and his GA Senate term ending Jan 2027 means a 2026 re-election win would make him the cycle's proven battleground winner." A 14.5-cent price is a big bet that all of that goes right, 27 months early.

Where The Panel Changed Its Mind

After round one, each seat read the other five seats' full reasoning, anonymized, and was allowed to revise. This is the Delphi method, and real forecasting panels use it for a simple reason: independent estimates protect you from groupthink, but a second look at your peers' evidence catches the things you missed. The instructions explicitly warn that drifting toward the average for its own sake is a failure, so what moved, and what refused to move, is signal.

The best revision of the run was a model catching its own mistake. Opus had been the panel's biggest Harris skeptic at 8%, and raised her to 12 after re-checking a peer's objection against the card exposed a flaw in its own reasoning:

"I called Emerson one of the 'highest-rated' pollsters, but its 4.55 avg miss is among the card's worse ratings" — Claude Opus

Sonnet, the panel's Harris bull at 22, came down to 19.5 from the opposite direction: "Forecaster C's specific point that AtlasIntel — the single best-rated pollster on the card (2.21 avg miss, neutral lean) — inverts the whole board (AOC 26, Buttigieg 22, Newsom 21, Harris 13)." [Editor's note: by the card's graded records, TIPP Insights (2.15-point average miss) rates best in the file; AtlasIntel's 2.21 is second.] Fable conceded a different blind spot entirely: "Forecasters D and E cited the Michigan statewide poll where Harris leads at 25 — evidence I had effectively ignored while building my name-ID-mirage case entirely on New Hampshire."

And one seat pushed back on the pushback. Kimi held its Harris number and flagged that a peer's argument had misfiled the evidence, noting that a "claim that the top-rated pollsters favor Buttigieg misreads Emerson (4.55, mid-pack)" when TIPP (2.15) and Echelon (2.93), two of the three best-graded records in the file, still show Harris leading. Net effect across the panel: Harris nearly unchanged at 16.3, Buttigieg up a point to 13.8, the field bucket up to 6.1, and Newsom, who nobody argued about, drifting up to 16.9 and the top of the board.

Dark Horses The Panel Won't Dismiss

Thirty-plus names on this board blend under 1%, but the seats were asked which longshots they would least want to dismiss, and three kept coming back. Wes Moore (2.5 cents, 2.1% blend) was named by five of the six seats, always on the same structural logic: a sitting governor with, in Opus's words, "the profile type that historically converts." Andy Beshear (3.8 cents, 3.6%) is the electability case; DeepSeek's version: "a Democratic governor reelected in deep-red Kentucky, his executive experience and crossover appeal could make him a compromise pick if the party seeks electability." And Zohran Mamdani (0.2 cents, 0.5%) is the wildcard platform play, in Sonnet's phrasing the "NYC mayor since 2026 gives an outsized media platform despite total lack of national primary history." At two-tenths of a cent, that argument is the entire question.

Every Seat's Number

The table below shows each seat's final board, after the revision round, for every candidate blended at 1% or higher; the complete 47-row board with every name is up top. Values are percentages; the Kalshi column is the verdict-day print (the live, daily-refreshed prices are in The Board above).

Candidate	Kalshi (verdict day)	Fable	Opus	Sonnet	GLM	Kimi	DeepSeek	Blend
Gavin Newsom	16.5c	17.5	15.0	17.4	17.0	17.0	17.5	16.9%
Kamala Harris	9.4c	12.4	12.0	19.5	20.5	15.0	18.6	16.3%
Pete Buttigieg	5.5c	12.4	13.0	14.4	14.8	12.0	16.5	13.8%
Alexandria Ocasio-Cortez	13.5c	9.3	10.0	10.3	10.2	9.0	9.3	9.7%
Josh Shapiro	4.5c	6.2	6.0	6.2	6.8	6.0	8.2	6.6%
The Field (anyone unlisted)	not offered	6.2	7.0	7.4	6.8	3.0	6.2	6.1%
Mark Kelly	6.2c	3.1	4.0	4.1	5.7	5.0	3.1	4.2%
Andy Beshear	3.8c	4.1	3.0	3.1	3.4	4.0	4.1	3.6%
Gretchen Whitmer	1.0c	4.1	3.0	2.6	3.4	2.0	2.1	2.9%
Jon Ossoff	14.5c	3.1	3.0	1.5	2.3	3.0	2.1	2.5%
Wes Moore	2.5c	2.1	3.0	2.1	2.3	2.0	1.0	2.1%
Cory Booker	0.7c	2.1	2.0	1.3	2.3	2.0	2.1	2.0%
J.B. Pritzker	2.6c	4.1	3.0	—	—	3.0	—	1.7%
Bernie Sanders	0.7c	—	0.3	1.0	2.3	1.0	2.1	1.1%
Tim Walz	0.1c	1.0	1.0	0.3	2.3	1.0	1.0	1.1%

Every seat on this panel is graded against real market settlements — records to date: GLM 80% on 1,874 graded calls · Kimi 84% on 1,863 graded calls · DeepSeek 80% on 1,877 graded calls. Recomputed daily; the full scoreboard is public.

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.

More live boards from this panel: 2028 Republican nominee · J.D. Vance 39.5c · Marco Rubio 29.5c — and the 2028 presidential election board both primaries feed into.

Reading These Prices Like A Bettor

Take the board's strangest number and walk it through. Jon Ossoff's Yes trades at 14.5 cents, and a Kalshi price in cents is just the crowd's probability with a dollar sign on it: 14.5 cents means roughly a 14.5% chance, which converts to about +590 in sportsbook odds. One wrinkle before you compare that to the panel: the board's 46 midpoints sum to about 100.7 cents, not 100. That extra 0.7 plays the role a sportsbook's vig plays, a small margin baked into the sum of the prices (on an exchange it comes from bid-ask spreads rather than a bookmaker), and stripping it out, what bettors call de-vigging, leaves Ossoff's true market-implied chance at about 14.4%. The panel's 2.5% blend, translated the same way, prices the same contract in the neighborhood of +3900. That is the size of the disagreement in a bettor's terms.

Now the honesty a sportsbook line never forces on you. This market's settlement is as much as 27 months away, so anyone taking a position is locking money up until the 2028 convention, paying trading fees on the way in, and on the sub-cent names, crossing spreads nearly as wide as the price itself. Most positions on a board like this resolve through line movement long before settlement, not at it. The full cents-to-odds mechanics are in our sportsbook-to-Kalshi translation guide, and if exchange boards are newer to you than sportsbooks, start with our guide to Kalshi's politics markets.

What Would Change The Panel's Mind

Four concrete things, each checkable, each with a direction:

• A new print from a top-graded pollster. AtlasIntel and TIPP anchor opposite sides of the Harris argument with data from May. A fresh wave from either one moves her number the way it points, and probably by a lot.
• November 2026. Ossoff's Georgia reelection is the gate on his entire national story: a win validates the market's 14.5 cents, a loss ends the trade. Several seats also treat the midterms as the audition tape for the governor tier.
• More early-state polling. The panel's Buttigieg conviction rests heavily on two New Hampshire polls and one Michigan poll. Every new early-state survey either compounds that case or breaks it.
• Campaign infrastructure. Staff hires, PAC filings, and early-state visits are the things name-recognition polling cannot see. The first candidate to show real machinery, the models repeatedly note, converts soft support into the kind that survives a primary calendar.

Settlement Timeline

Settles	At the 2028 Democratic National Convention, when a nominee is chosen and accepts (listed expiration November 7, 2028)
Next known catalysts	November 2026 midterms (Ossoff's Georgia race, the governor tier's report cards) · the DNC's 2028 primary calendar decisions · declared candidacies through 2027
This page	Prices in the board above refresh daily from Kalshi's live order book; the panel re-scores when the story materially moves, with every version graded in public

The Rest Of The 2028 Cluster

This board is one third of the exchange's 2028 picture: the 2028 Republican nominee board runs the same protocol on the other primary, and the 2028 presidential election board prices the general-election field both feed into. Every other market this panel has priced, from championships to crypto, lives on the prediction markets hub.

The Shape Of The Distribution

Step back from the names and the two boards disagree about geometry, not just order. The market concentrates its probability in three names (Newsom, Ossoff, Ocasio-Cortez at a combined 44.5 cents); the panel spreads the same weight across four different ones (Newsom, Harris, Buttigieg, Ocasio-Cortez at 56.7 points) and pushes the rest down into the governors' tier and the unlisted field. But the two boards make one claim together: a favorite at 16-17% this far from a convention is a statement that this race is more open than any single story about it.

To be explicit about what this page is: an analytical comparison of one crowd's prices and one panel's estimates, on a CFTC-regulated exchange available to adults 18+ in eligible states as of August 2026. It is not advocacy for any candidate or party, and none of it is advice about what anyone should trade or how anyone should vote. The panel's numbers will be graded against settlement in 2028, the same way every board it prices is graded.

Verdicts generated August 3, 2026, in two rounds (independent, then revision). Prices as of August 3, 2026. Panel: Claude Fable, Claude Opus, Claude Sonnet, GLM 5.2, Kimi K3, DeepSeek V4, all price-blind. These are model estimates, not predictions of fact and not financial advice.

Prices on this page are Kalshi's book. If you also trade on Polymarket, code OS4 gets new users a $20 bonus on a $10 deposit — affiliate link; terms as stated by Polymarket; 18+, availability varies by state.