TL;DR
The panel's top of the board: Arch Manning (11%), Julian Sayin (7%), Jeremiah Smith (5%) (ties shown at equal blends in the table). Every verdict below was made price-blind, and every one gets graded against the real settlement on our scoreboard.
The Heisman is the most base-rate-driven award in sports: quarterbacks on playoff-caliber teams win it most years. Kalshi prices 28 names, led by Notre Dame's CJ Carr and Texas's Arch Manning, and the models' job is mostly to decide which quarterbacks sit on the teams that will actually be in the playoff conversation in November.
What The Models Were Given
The models ran knowledge-only with an anti-misattribution instruction (no roster ledger mid-offseason), disclosed here per our data-card rules. The market names themselves come from Kalshi's live board, so every candidate listed is a real contract trading now. Eight model seats were asked on every outcome; where a seat's verdict did not return, its table cell shows an em dash and the blend averages the seats that answered.
Where The Machines Split From The Money
CJ Carr: market 13¢, AI blend 3% (10 points below the crowd). Gemini 3.1 Pro: "While Carr is a highly-touted talent at a major program, predicting any specific Heisman winner involves overcoming long odds and uncertainty regarding starting status."
Trinidad Chambliss: market 7¢, AI blend 1.1% (6 points below the crowd). Gemini 3.1 Pro: "As an obscure player, Chambliss faces near-impossible odds in an award heavily dominated by star quarterbacks on top-tier Power Four championship contenders." [Editor's note: this model's premise is wrong — Chambliss is Ole Miss's starting quarterback and a widely covered contender, not an obscure player. The verdict stands as published and will be graded as such.]
Darian Mensah: market 7¢, AI blend 1.2% (6 points below the crowd). Gemini 3.1 Pro: "As a quarterback likely outside the elite tier of championship-contending programs, his baseline probability of winning the Heisman Trophy is exceptionally low."
Josh Hoover: market 6¢, AI blend 1.0% (5 points below the crowd). DeepSeek V4: "Josh Hoover is a solid but unspectacular QB at TCU, a program not expected to contend for a playoff spot, and Heisman winners typically come from top-5 teams." [Editor's note: this model's premise is wrong — Hoover transferred from TCU to Indiana in January 2026; the TCU attribution is stale. The verdict stands as published and will be graded as such.]
The Full Board
| Outcome | Kalshi | AI blend | ChatGPT (GPT-5.5) | Claude Fable | Claude Opus | Claude Sonnet | Gemini 3.1 Pro | GLM 5.2 | Kimi K3 | DeepSeek V4 |
|---|---|---|---|---|---|---|---|---|---|---|
| Arch Manning | 12¢ | 11% | 12% | 13% | 9% | 15% | 8% | 15% | 12% | 2% |
| Julian Sayin | 7¢ | 7% | 9% | 12% | 12% | 6% | 5% | 4% | 7% | 3% |
| Jeremiah Smith | 5¢ | 5% | 6% | 7% | 6% | 12% | 1.5% | 3% | 6% | 1.2% |
| Bryce Underwood | 2¢ | 4% | 6% | 7% | 6% | 6% | 2% | 4% | — | 1.0% |
| Dante Moore | 6¢ | 4% | 4% | 9% | 5% | 1.5% | 2% | 2% | 6% | 0.5% |
| CJ Carr | 13¢ | 3% | 2% | 5% | 5% | 3% | 1.5% | 2% | 5% | 2% |
| LaNorris Sellers | 2¢ | 3% | 1.8% | 6% | 4% | 5% | 1.5% | 3% | — | 0.5% |
| John Mateer | 7¢ | 3% | 4% | 4% | 5% | 2% | 0.5% | 4% | 4% | 2% |
| Keelon Russell | 3¢ | 3% | 4% | 5% | 5% | 1.0% | 1.2% | 4% | — | 1.5% |
| Gunner Stockton | 4¢ | 3% | 4% | 2% | 2% | 5% | 1.2% | 3% | 4% | 0.5% |
| Sam Leavitt | 3¢ | 3% | 2% | 4% | 3% | 5% | 0.5% | 3% | — | 0.5% |
| Marcel Reed | 5¢ | 2% | 2% | 4% | 4% | 1.0% | 1.5% | 1.5% | 3% | 0.5% |
| Jayden Maiava | 4¢ | 1.9% | 4% | 1.5% | 1.2% | 5% | 1.2% | 0.5% | — | 0.5% |
| Demond Williams Jr. | 2¢ | 1.5% | 1.8% | 2% | 2% | 1.0% | 1.5% | 1.0% | — | 0.5% |
| Darian Mensah | 7¢ | 1.2% | 0.5% | 2% | 1.8% | 1.5% | 0.2% | 1.5% | 1.2% | 0.5% |
| Ahmad Hardy | 1¢ | 1.2% | 0.4% | 2% | 3% | 2% | 0.1% | 0.5% | — | 0.1% |
| Austin Mack | 1¢ | 1.1% | 1.2% | 2% | 0.8% | 1.0% | 1.5% | 1.5% | 0.7% | 0.5% |
| Kevin Jennings | 3¢ | 1.1% | 0.6% | 1.5% | 1.2% | 1.0% | 1.0% | 2% | — | 0.5% |
| Trinidad Chambliss | 7¢ | 1.1% | 0.2% | 4% | 2% | 0.5% | 0.1% | 1.0% | 0.8% | 0.3% |
| Jaron-Keawe Sagapolutele | 2¢ | 1.0% | 0.3% | 4% | 0.8% | 1.0% | 0.2% | 0.5% | — | 0.1% |
| Josh Hoover | 6¢ | 1.0% | 1.2% | 1.2% | 0.8% | 1.0% | 1.5% | 1.2% | 0.6% | 0.3% |
| Rocco Becht | 2¢ | 1.0% | 1.2% | 0.8% | 1.2% | 1.0% | 0.5% | 1.5% | — | 0.5% |
| Kewan Lacy | 1¢ | 0.9% | 0.2% | 1.5% | 1.2% | 1.5% | 0.1% | 1.5% | — | 0.2% |
| Brendan Sorsby | 1¢ | 0.9% | 0.2% | 2% | 2% | 1.0% | 0.4% | 0.4% | — | 0.1% |
| Bo Jackson | 1¢ | 0.8% | 0.3% | 2% | 1.2% | 0.2% | 0.2% | 1.0% | — | 0.1% |
| Malachi Toney | 3¢ | 0.7% | 0.4% | 2% | 1.2% | 0.5% | 0.1% | 0.5% | — | 0.1% |
| Byrum Brown | 2¢ | 0.6% | 0.3% | 0.5% | 0.8% | 1.0% | 0.5% | 0.5% | — | 0.5% |
| Drew Mestemaker | 2¢ | 0.5% | 0.1% | 0.6% | 0.8% | 0.5% | 0.1% | 1.0% | — | 0.1% |
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.
What data did the models see?
The models ran knowledge-only with an anti-misattribution instruction (no roster ledger mid-offseason), disclosed here per our data-card rules. The market names themselves come from Kalshi's live board, so every candidate listed is a real contract trading now.
