Jared Moskowitz won the FL-25 Democratic primary on Tuesday, August 18, 2026, beating Democratic Socialist Oliver Larkin roughly 63 percent to 36 percent. Kalshi's market is finalized: the Moskowitz contract settled at $1.00, Larkin's at zero. This page, which tracked the race live, is now the graded record of what the market and our price-blind AI panel said before the vote, and how each call held up.
Updated August 19, 2026 · 8 min read · By Sam Smith
Redistricting scrambled South Florida this cycle, and FL-25 is where the reshuffle was clearest. The seat's longtime incumbent, Debbie Wasserman Schultz, drew new lines that pushed her into a different district, and a sitting congressman from next door, Jared Moskowitz, moved in to run for it. Kalshi listed a separate yes/no contract for each candidate, and before the vote we asked six AI models to score the whole field without letting any of them see a single price. The result is in, the contracts have settled, and the honest part of this exercise starts now: grading what everyone said against what actually happened. The scorecard, including the one number where Larkin beat his polls outright, is below.
The Quick Answer
Jared Moskowitz wonthe FL-25 Democratic primary on August 18, 2026, roughly 63 percent to 36 percent over Oliver Larkin with 95 percent of precincts reporting. Decision Desk HQ projected the race at 7:08 p.m. ET, minutes after polls closed. Kalshi's contracts are finalized: Moskowitz settled at $1.00 after trading near 94¢ on primary-day morning, Larkin settled at zero from 6¢. Our price-blind six-model panel, frozen on August 12, blended Moskowitz to about 89%, right side, a nickel less certain than the crowd. The full grade, including the one thing the panel's 10% Larkin tail got half-right, is below.
- The Result: Moskowitz ~63%, Larkin ~36% (95% of precincts reporting), called minutes after polls closed
- The Settlement: Moskowitz resolved YES at $1.00 (last trade 99¢); Larkin and Wasserman Schultz resolved NO at zero
- The Grade: the market showed ~94¢ on the winner the morning of the vote, the panel's frozen blend said ~89%; both right, the crowd righter
- What's Next: Moskowitz faces Republican Scott Singer on November 3 in a district both Cook and Sabato rate a toss-up
Stay ahead of the markets.
Daily insights and expert picks on Kalshi, Polymarket, and what's moving markets.
Free forever. Unsubscribe anytime.
New to Kalshi event markets? Start with our plain-English guide to how Kalshi contracts actually work →
These are market prices and model estimates, not predictions of fact and not financial advice.
The Market At A Glance
| Venue | Kalshi |
| The Event | "FL-25 Democratic nominee?" (Kalshi event KXFLPRIMARY-25D26), one yes/no contract per candidate |
| Status | Settled. Every contract is finalized: Moskowitz resolved YES at $1.00; Larkin and Wasserman Schultz resolved NO at zero |
| How It Resolved | Moskowitz won the August 18 primary, roughly 63-36 with 95 percent of precincts reporting; Florida primaries are plurality, no runoff |
| When Trading Closed | The evening of August 18, 2026, once the nomination was decided. The far-off expiration timestamp listed on the contract was always a placeholder |
| Prices On This Page | The board below is the frozen pre-vote record, the morning-of-primary prices (Moskowitz ~94¢, Larkin ~6¢) and the panel's August 12 verdicts, preserved so the grade is checkable |
These are model estimates, not predictions of fact and not financial or voting advice, now graded against a settled result. Kalshi event contracts trade on a CFTC-regulated exchange; you must be 18 or older and in an eligible state to participate, and this article does not advocate for or against any candidate.
How Redistricting Set This Race
The number that explains this primary was never a poll, it was a map. Four fetched facts defined the board:
- The Old Incumbent Left FL-25. Debbie Wasserman Schultz represented FL-25, but Florida's 2026 GOP-led mid-decade redistricting redrew South Florida. Under the new lines, Wasserman Schultz opted to run in FL-20, a safe Democratic seat. She was not on the FL-25 ballot.
- A Sitting Congressman Moved In. Jared Moskowitz has represented FL-23 since 2023, but the redraw dismantled his district, so he announced he would run in the new FL-25, which stretches along the coast from Delray Beach to Miami Beach.
- The General-Election Math Changed Too. Per the Cook Political Report's race analysis, the redraw shifted FL-25 from a seat Kamala Harris carried by about 5 points in 2024 to one Donald Trump would have won by 9, and both Cook and Sabato's Crystal Ball rate the November race a toss-up. That fight is now live: Moskowitz versus Republican primary winner Scott Singer.
- The Primary Field Was Two Candidates. On the Democratic side it was Moskowitz versus Oliver Larkin, a first-time candidate and member of the Democratic Socialists of America.
The Final Board
The settled scorecard: what each contract did against the frozen pre-vote prices and the panel's price-blind blend.
| Candidate | Frozen market (Aug 18 AM) | AI blend (Aug 12) | Result |
|---|---|---|---|
| Jared Moskowitz | ~94¢ | 89.3% | WON ~63% · resolved YES at $1.00 |
| Oliver Larkin | ~6¢ | 9.8% | Lost, ~36% · resolved NO |
| Debbie Wasserman Schultz | ~1¢ | ~0% | Not on the ballot · resolved NO |
These are model estimates graded against a settled result, not financial or voting advice. The market column is the morning-of-primary price; the AI blend is the simple mean of the six model numbers taken August 12, price-blind, after a revision round. The row worth sitting with is Larkin's: a contract that settled at zero, exactly as both the crowd and the panel expected, attached to a candidate who nearly doubled his polling topline in actual votes. That gap between a lost contract and a real overperformance is the most instructive thing on this board, and we grade it fully below.
More live boards from the same panel: Florida Republican governor primary (the same August 18 Florida ballot, statewide) · Wisconsin Democratic governor primary (David Crowley, settled YES at 100¢ after an election-night reversal) · Minnesota Republican governor primary (Lisa Demuth, settled YES at 100¢).
Every Seat's Number (Price-Blind)
[Editor's note: this table is the frozen pre-vote record, preserved for grading. The six model verdicts were generated price-blind on August 12, 2026, before the August 18 primary. Each model saw a fetched, dated summary of the candidates, the redistricting timeline, the fundraising gap, and every public poll, but no market prices. The "Market" column is the primary-day price the panel was blind to. Moskowitz won; his contract resolved YES.]
| Candidate | Market | Fable | Opus | Sonnet | GLM | Kimi | DeepSeek | Blend | Gap (blend% − mkt¢) |
|---|---|---|---|---|---|---|---|---|---|
| Jared Moskowitz | 94¢ | 90.0 | 89.0 | 91.0 | 90.0 | 86.0 | 90.0 | 89.3% | -5 |
| Oliver Larkin | 6¢ | 9.0 | 10.0 | 8.0 | 10.0 | 13.0 | 9.0 | 9.8% | +4 |
| Debbie Wasserman Schultz | 1¢ | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | ~0% | -1 |
These are model estimates graded against a settled result, not financial or voting advice. The Blend column is the simple mean of the six model numbers in its row, taken after a revision round in which each model read the others' anonymized reasoning. The last column is the directional gap the panel carried into the vote: five cents shy of the market on the winner, four points heavier on the challenger's tail. The result cashed the market's side of both gaps.
A note on the panel. Our standard Model Verdict lineup runs eight seats. On this run two were unavailable (the ChatGPT and Gemini seats were not installed on the machine that produced these verdicts), so this board ran on six models: Claude Fable, Claude Opus, Claude Sonnet, GLM, Kimi, and DeepSeek. All six scored the field on August 12, 2026 without seeing any prices. Each model's number lands in our back-tested Model Verdict ledger, where past price-blind calls are graded against settled results, this page now being one of them, so the method is measured, not just asserted.
Jared Moskowitz: Why 94 Cents Was The Right Price
Every model made Moskowitz the favorite, and none put him below 86%. The case rested on three fetched facts that all pointed the same way, and every one of them held on election night. First, the public polling: Beacon Insights' July 22 wave (n=532 likely Democratic primary voters, margin of error ±3.6) had Moskowitz at 55.6% to Larkin's 19.9%, with 24.5% undecided. Second, the money: Moskowitz raised roughly $2.5 million to Larkin's roughly $550,000 and carried an 11-to-1 cash lead into the final 10 days. Third, and heaviest in the models' reasoning, the banked vote: among the 21% of respondents who had already returned mail ballots, Moskowitz led 75.3% to 19.3%.
That last number was the whole race in miniature. As DeepSeek put it before the vote, the mail-vote lead "is the most concrete data point and creates a massive hurdle for Larkin." Opus reduced it to arithmetic: with a fifth of the electorate already voting 75-19, "Larkin must win the remaining electorate by an implausible margin even if the 24.5% undecided break heavily his way." The undecideds did break heavily Larkin's way, as it turned out, and the arithmetic still wasn't close. Decision Desk HQ projected the race at 7:08 p.m. ET, minutes after polls closed, with the margin running about 63-36 at 95 percent of precincts reporting. "Tonight, Democratic voters sent a clear message. I faced a challenge from the DSA. They threw everything at me and I didn't change who I am or what I believe, and the voters overwhelmingly chose the independent-minded approach I've brought to Congress," Moskowitz said in his victory remarks.
Oliver Larkin: The Tail That Didn't Cash, And The Number That Did
The panel refused to price Larkin to zero, holding him near 10% against the market's 6¢, and that disagreement is the most interesting thing to grade. Larkin ran a Democratic Socialist field operation built on more than 30,000 doors knocked, roughly 100,000 calls and texts, and 70-plus volunteers. Fable kept Moskowitz at 90 rather than 95 specifically because that field program is "the classic mechanism for polling misses in low-turnout primaries." Kimi, the high seat on Larkin at 13%, flagged the campaign's internal poll claiming a tie: "if true, massive polling miss, but no independent confirmation and campaign internals average +6 bias."
Here is the honest grade. As a win probability, the tail was wrong, and the market's stingier 6¢ was the better number: Larkin's contract settled at zero and the race was called within minutes of polls closing. But the mechanism the tail priced, an undercounted insurgent electorate breaking late, showed up in the vote share. Larkin finished near 36 percent, nearly double his 19.9 percent polling topline, meaning the undecideds broke overwhelmingly his way, exactly the pattern Sonnet had in mind citing "DSA-style grassroots campaigns occasionally overperform polls (AOC 2018, Bowman 2020)." What saved the favorite was the fact every model leaned on hardest: the 75-19 banked mail vote meant a huge chunk of the electorate had already voted before any late break could touch it. A field program can move undecideds; it cannot move ballots already returned. "The fight continues," Larkin told supporters on election night. In this district, against this arithmetic, it never really started.
Where The Panel Changed Its Mind
After round one, every model read the other seats' anonymized numbers and reasoning and could revise. The movement was small but pointed in one direction, toward the banked-vote evidence, and the result vindicated the direction of every move.
- GLM Moved Up, And Said Why. It lifted Moskowitz from 82% to 90% after reading the group, concluding that "the peer consensus reflects this reality, correctly weighting the early vote over the raw top-line poll." — GLM
- Fable Moved Down, Toward The Tail. It trimmed Moskowitz from 93% to 90%, giving Larkin's ground game more room: it discounted "for the 24.5% undecided share, a single-pollster text-methodology risk, and Larkin's ... large field program (30K doors) ... that keeps me at 90 rather than 95." — Fable
- Sonnet Held At 91%, On Purpose. It saw no reason to move because "independent polling didn't just hold but widened over two waves," and the mail-vote subsample "undercuts any 'quiet momentum' story for Larkin." — Sonnet
The net effect was convergence: the six seats started spread from 81% to 93% on Moskowitz and finished bunched from 86% to 91%, landing the blend at 89.3%, a nickel under the 94¢ the market showed on primary-day morning. Note the two revisions that pulled in opposite directions: GLM's move toward the banked vote graded cleanly, while Fable's trim toward the tail bought insurance the result never claimed, though the undecided break it feared was real in the margin.
Grading The Panel Against The Result
This is the section this page existed for. Every Model Verdict board carries the same promise: the numbers get graded in public once the market settles. The FL-25 scorecard, line by line:
- The Market: Right, And Efficiently SO. Kalshi priced Moskowitz as a prohibitive favorite from the panel's August 12 run onward, showed him near 94¢ the morning of the vote, and the last trade printed at 99¢ before the contract finalized at $1.00. On a two-candidate board that summed to about 100 with no overround to strip, 94¢ meant 94 percent, and the favorite won going away. No notes.
- The Panel: Right Side, Five Cents Shy. The blend's 89.3% was directionally correct and inside every seat's stated confidence, but the market's extra nickel of conviction was earned. Scored the way our ledger scores everything, distance from the settled outcome, the crowd's primary-day price beat the panel's frozen number on both live contracts.
- The Tail: Wrong As A Price, Half-Right As A Story. The panel's 9.8% on Larkin against the market's 6¢ was the board's one real disagreement. The contract settled at zero, so the market wins the grade, but Larkin's ~36 percent against a 19.9 percent topline says the "field program moves undecideds" mechanism the models paid for was real, just aimed at vote share instead of victory.
- The Zero: Exactly Zero. All six seats put Wasserman Schultz at 0.0% while her contract floated at a cent; it resolved NO, as a non-candidate's contract must. The panel's cleanest line.
One more grading note for readers of the Florida Republican governor primary board: the same August 18 ballot settled that market's statewide question too, and the same public-grading treatment applies there. Every settled call, this page included, rolls up on the graded scoreboard.
The Traps: Rows That Looked Like Signal But Were Not
Transparency matters more than a clean story, so here is where the table would have misled you, past tense now, each one resolved the way this page said it would:
- The Wasserman Schultz Contract Could Never Win. She was the former FL-25 incumbent, but redistricting moved her, and she ran in FL-20, not FL-25. Her contract sat near the one-cent floor all cycle and resolved NO at settlement. Do not read a listed price as candidacy.
- The Internal "Tie" Poll Was Not Independent. Larkin's campaign released a poll claiming a close race. The final 27-point margin, closer than the public polling gap but nowhere near a tie, split the difference in the least flattering way for the internal: Larkin outran his public topline by 16 points and still lost by 27.
- The Price Was Softer Than It Looked, Until It Wasn't. This market traded thin for most of its run before volume surged into primary day, roughly 950,000 contracts across the two live markets by the morning of the vote. The lesson stands even though the number held: on a thin board, treat any single quote as a range.
Settlement Timeline
| The Vote | Tuesday, August 18, 2026 (Florida primary; plurality, no runoff). Moskowitz won roughly 63-36 with 95 percent of precincts reporting |
| The Call | Decision Desk HQ projected the race at 7:08 p.m. ET; NBC News also called it on election night |
| Kalshi Settlement | Trading closed the evening of August 18 and every contract is finalized: Moskowitz YES at $1.00, Larkin NO, Wasserman Schultz NO. The far-off expiration timestamp listed on the contract was a placeholder, not the decision date |
| Next Catalyst | The general election is November 3, 2026: Moskowitz versus Republican Scott Singer in a district both Cook and Sabato rate a toss-up under the new lines |
| Re-Score Policy | This board is settled and graded; the pre-vote prices and panel verdicts above are frozen as the permanent record. Any November general-election market gets its own board |
How To Read A Settled Price
The models told you where they parted from the crowd; the settled board is where the habit of reading prices carefully pays off. Work the example one last time. Moskowitz's contract showed 94¢ on primary-day morning. On a crowded field you would strip the board's overround by summing every candidate's YES price, but here the two live contracts (94¢ and 6¢) already summed to about 100, so 94¢ mapped cleanly to roughly 94 percent, and the 94 percent shot came in. That is what an efficient two-horse board looks like, and checking it is the whole idea behind converting Kalshi's cents into real odds before comparing anything.
That same discipline of comparing a number to its fair value is the method behind every board we run, Senate seats and governors' races alike, collected on our 2026 Senate election odds hub. And the method's full record, every settled call scored, lives on the graded scoreboard.
This board lived on Kalshi and our panel does not trade. But if you came for the method rather than this one race, OddsShopper's free expert picks today apply the same shop-the-number, find-the-fair-value discipline to the live sports slate.
A few honest points before reading too much into a settled scorecard:
- Eligibility. Kalshi is a CFTC-regulated event-contract exchange, not a sportsbook. You must be 18 or older and located in an eligible state, and availability varies by state and can change.
- One Race Is One Data Point. The market beat the panel by a nickel here; the ledger, not any single board, is where the method is actually measured.
- Results Context. Vote figures on this page are near-complete unofficial returns (95 percent of precincts reporting as called on election night), quoted from the cited coverage; Florida's official certification follows later.
- Not Advice. Everything here is analysis for information, not a recommendation to trade or to vote. These are model estimates graded against a settled result, not predictions of fact and not financial or voting advice.
The Bottom Line
This was the rare board where the crowd and the models mostly nodded at each other, and the vote confirmed them both. Redistricting cleared the FL-25 incumbent out, a well-funded moderate congressman moved in, and the banked mail vote every model leaned on, 75-19 before election day, did exactly what banked votes do: it made the late break irrelevant. Moskowitz won roughly 63-36, his Kalshi contract settled at $1.00, and the market's morning-of 94¢ graded a nickel better than the panel's frozen 89. The one thing worth remembering from the losing side is the shape of Larkin's number, a 36 percent finish against a 19.9 percent topline, because "the field program moves undecideds" turned out to be true and insufficient at the same time. The story now moves to November 3, Moskowitz versus Scott Singer in a district rated a toss-up by both major handicappers, and when a market prices that race, it gets its own board. If this method is your thing, our guide to Kalshi versus Polymarket is the natural next read.
More on this: Why Chris Pappas (D) Will Win the NH Senate Primary · Why Angie Craig (D) Will Win The Minnesota Senate Primary · Why Billy Benavidez Will Win Wyoming's Democratic Senate Primary · Why Jack Reed Will Win The Rhode Island Democratic Senate Primary · Why N'Kiyla Thomas Will Win Oklahoma's Democratic Senate Primary
Hottest Prediction Markets Right Now
- 2028 Democratic presidential nominee · $166M traded
- Rocket Classic Winner · $124M traded
- 2027 Pro Football Champion · $51M traded
- Pro Baseball Champion · $40M traded
- Bitcoin price at the end of 2026 · $30M traded
- 2027 Pro Basketball Champion · $19M 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.
Prices on this page are Kalshi's book. If you also trade on Polymarket, code OS4 gets new users a $50 trading bonus on a $10 deposit — affiliate link; terms as stated by Polymarket; 18+, availability varies by state.
