The cross-venue spread is the gap between the probability Polymarket and the probability Kalshi assign to the same AI release happening by the same date — and you read it as a confidence and freshness indicator, not as an arbitrage. A tight spread means both crowds agree; a widening spread means new information, a resolution-rule mismatch, or both. Next AI Drop blends the two into a single odds input rather than trusting either alone.
What is the cross-venue spread?
The cross-venue spread is the difference, in percentage points, between Polymarket's and Kalshi's implied probability for the same AI release event. If Polymarket prices "model X ships by June 30" at 62% and Kalshi prices the comparable contract at 57%, the spread is five points. In normal conditions the gap stays small — typically inside a handful of points — because both venues watch the same public evidence and arbitrage-minded traders keep obvious mispricings from lasting. We define it more formally in the cross-venue spread defined entry; this post is the how-to.
Treat the spread as a derived metric, not a price you trade. A small, stable spread tells you the forecast is well-formed. A spread that moves is the part worth reading.
Why does the same model get two prices?
The same model gets two prices because Polymarket and Kalshi are different venues with different rules, traders, and liquidity — so identical headline questions are not actually identical contracts. Both are sources we build on, and the divergence between them is exactly what makes running both worthwhile. Four structural causes do most of the work.
Different resolution criteria
The biggest single cause is resolution wording. One venue may resolve a market on "the company releases a model in the family by date" while the other resolves on "a specific named model is generally available by date." Those are not the same bet. A company can ship a smaller sibling — say ⟨EMBER-ALPHA⟩ — that satisfies the company-level contract but not the specific-model one. That single difference can open a multi-point gap with no new information at all, which is why the first thing to check on any spread is whether you are comparing like with like.
Different trader bases and liquidity
The two venues draw different crowds. Kalshi is a regulated US exchange; Polymarket pulls a more global, crypto-native base. Thinner liquidity on one side means a single sized order moves its price further, so a low-volume contract can drift from its better-capitalised twin without reflecting any change in the underlying odds. Volume is precisely why our score weights it — see how the blended odds feed Drop Readiness.
Information arriving on one venue first
New evidence — a changelog edit, an official post, a public catalog listing — rarely hits both order books at the same instant. When a signal lands, the venue with faster or more attentive traders re-prices first, and the spread widens until the other catches up. That lag is short-lived, but while it lasts the spread is pointing at fresh information before it is fully priced everywhere.
How do you read a widening spread?
A sustained widening spread usually precedes a re-rating: it flags that new information has reached one venue and not yet propagated to the other, or that the two contracts are resolving on genuinely different events. A brief one-off divergence is noise; a gap that holds or grows over hours is the signal. When you see it, do not average blindly — first confirm both contracts cover the same event, then look for the intel that explains the move.
Illustrative example. Suppose a Claude-class release is priced at 71% on one venue and 59% on the other, a 12-point gap that holds for a day. That divergence reads as either a resolution mismatch (one contract counts any model in the family, the other a specific checkpoint) or fresh intel that has hit one crowd first. The next confirmed post or catalog change typically collapses the gap toward the better-informed side. These numbers are illustrative, not live odds — the live figures live on the home timeline. For how to weigh that intel against the market move, see reading market moves as signal.
Why use both venues instead of one?
We use both venues because blending Polymarket and Kalshi cancels the single-venue noise that creates spreads in the first place — thin liquidity, a lopsided trader base, or a slow re-price on one side. A two-venue blend is more robust than either price alone, and that blended figure is what carries the heaviest weight in our score: odds account for 45% of Drop Readiness (DR = .45 odds + .25 intel + .20 deadline + .10 volume). Polymarket and Kalshi are the sources we build on; we add the second venue, hourly tagged intel, and the readiness math on top. The full computation lives at how the blended odds feed Drop Readiness, and the product-level contrast — what each venue covers and where they complement each other — is laid out on Polymarket and Kalshi compared.
If prediction markets as a forecasting tool are new to you, start with how prediction-market odds work before applying the spread read here.
Signal, not stakes
The cross-venue spread is a forecasting read, not a trade recommendation. Next AI Drop is not a betting site and offers no financial or investment advice — we read Polymarket and Kalshi odds as the fastest public signal of when a model ships, nothing more. We have no affiliation with either venue. For how we source and cite every figure, see our odds sources.