Kalshi vs Polymarket fees in 2026: who's cheaper?

What does Kalshi charge?

Kalshi's published fee is a taker fee. Makers are listed at 0.0, and takers pay on entry and exit. The economic shape is simple: 0.07 x P x (1 - P) per contract per side, where P is the contract price in dollars. A 50c contract has P = 0.50, so the unrounded fee is 0.07 x 0.50 x 0.50 = $0.0175, or 1.75c. The fee gets smaller as the contract moves toward either end of the board.

Kalshi also publishes the fee with a ceiling function: fee = ceil(0.07 x C x P x (1 - P)) cents, where C is the number of contracts. That order-level rounding matters most on small tickets. The main comparison still starts with the same curve: the highest fee is near 50c, not near 10c or 90c. For live cross-venue context, the gap board uses Kalshi's 0.07 rate in its displayed fee-adjusted spread math.

What does Polymarket charge?

Polymarket also lists maker fees at 0.0 and taker fees as the relevant charge. The difference is that Polymarket's rate is not one flat number across the board. Its formula is C x rate x P x (1 - P), with the rate set by market category. That is the point most stale fee pages miss.

Crypto is 0.07, which matches Kalshi's rate before rounding. Sports, culture, economics, other, and weather are 0.05. Politics, finance, mentions, and tech are 0.04. Geopolitics is 0.0. That means the same 50c price can cost 1.75c in a crypto market, 1.25c in a sports market, 1.00c in a politics market, or 0.00c in a geopolitics market.

The price curve is still the same P x (1 - P) curve. The fee is largest at 50c and lower at 10c, 25c, 75c, and 90c. The category rate decides how tall that curve is.

Which venue is cheaper at 10c, 25c, 50c, 75c, and 90c?

At the same price, Polymarket is cheaper than Kalshi in every category with a rate below 0.07, equal before rounding in crypto, and cheaper still in geopolitics. That statement is just arithmetic on the published rates. It is not a statement about available liquidity, bid-ask width, state access, or whether the market definitions match.

The table below uses one contract, one side, and the unrounded fee curve. Kalshi order-level rounding can change the charged cents on a small ticket because its published formula uses a ceiling function. Polymarket's cost changes by category, so the table shows four useful buckets: crypto at 0.07, the 0.05 categories, the 0.04 categories, and geopolitics at 0.00.

The practical bettor's answer is still line shopping. A fee edge of 0.50c at 50c does not beat a worse price by multiple cents. The fee table tells what the vig costs once the price is equal.

Why does the TapeXray board rank matches at 7% before pricing them lower?

The live board has two different jobs. It compares live prices across roughly 10,900 matched pairs, and it has to do that before it can afford to read any single market's published Polymarket category rate. Reading every pair's rate up front would overrun the board's memory budget, so the fee-adjusted gap line runs in two passes.

Pass one ranks and gates every match at 7%, the conservative maximum any Polymarket category publishes. Pass two re-prices only the handful of pairs that clear the same-bet gate at that market's own published rate - 0.03 to 0.07 by category, some categories fee-free - so a pair's final number can only rise, never fall, and no pair is ever admitted at a price its real category rate would have excluded. A pair the gate rejects, shown separately as a suspect pair, keeps the 7% maximum and carries no per-market rate at all.

The displayed fee-adjusted gap formula is: fee_adjusted_cents = |mid_k - mid_p| - 200 x (0.07 x Pk x (1 - Pk) + rp x Pp x (1 - Pp)). Pk and Pp are the Kalshi and Polymarket mid prices in dollars, and rp is 0.07 in pass one or the gated pair's own category rate in pass two. The 200 multiplier reflects entry plus exit in cents across both venues.

Fork math is different. A locked fork holds to resolution, so TapeXray subtracts one-way taker fees per leg rather than a round trip. The fork net uses Kalshi ceil(7 x P x (1 - P)) cents and Polymarket PM_TAKER_FEE_RATE x P x (1 - P) x 100 cents, priced off executable YES and NO quotes rather than mids.

How big is the tape behind the comparison?

The fee schedule is static, but the usefulness of a fee comparison depends on what is actually trading. TapeXray's archive tracked 236,027 markets across two venues as of the 2026-09-21 archive update. The catalog listed 94,209 active Kalshi markets and 138,833 active Polymarket markets. It also listed 24-hour volume of $79,251,622 on Kalshi and $31,756,150 on Polymarket.

Those numbers do not mean every market can be crossed cleanly. The cross-venue board matched markets by title and rejected suspect pairs. On the 2026-09-21 board snapshot, there were 100 suspect pairs rejected, 6 pairs gated total, 4 pairs inside the fee line, and 2 pairs positive after fees. That is the usual lesson: most days there is nothing here. Fees matter most when a real price is available on both sides and the definitions match.

For venue availability by state, use the state pages. For archived market pages, use the markets hub. For plain-English fee and odds primers, use the guides hub.

Questions people ask

Related reading

• Real Kalshi-Polymarket gaps right now in 2026Live count of real Kalshi-Polymarket gaps after fees, with pairs inside the fee line, suspect rejections, category mix, and the latest board timestamp.
• What -110 costs in 2026: vig vs market feesA -110 sportsbook line needs 52.38% wins and carries a 4.76% overround. Compare that with visible Kalshi and Polymarket taker fees before betting.
• 63c on Kalshi: odds and probability in 2026Convert Kalshi cents into implied probability, American odds and decimal odds, then see why a 63c YES needs about 64.63% after the taker fee.