Lowest Fee Crypto Futures Exchanges

The lowest published trading commission in this comparison is zero, offered by Paradex Retail orders and Lighter Standard accounts. Those are useful choices for someone placing ordinary manual orders in an eligible location. Neither result says that every trade costs nothing: the execution price, funding and movement of collateral still affect the amount left in your account. For a centralized derivatives account, Deribit Standard has the lowest regular two-taker commission among the confirmed profiles below, while Binance and OKX provide the straightforward USDT comparison.

This table compares entry-level accounts without token holdings, negotiated terms or temporary rewards. Maker and taker describe how a fill reaches the order book, not whether you clicked a button labelled Limit or Market. A limit order that immediately matches another order can pay the taker rate. The official Bybit schedule separates its regular futures rate from spot and special trading zones; we use the regular futures row here.

Read the account column before comparing percentages. A Paradex order submitted through its normal API is Pro by default; an interactive API order can receive Retail treatment under its documented limits. A Lighter Premium account has a different tariff from Standard. Putting the free rate beside every trading channel would produce a misleading result even though the headline itself came from an official website.

All calculated examples use a 10,000 USDT or dollar-equivalent position at entry and the same value at exit, unless stated otherwise. This keeps the comparison about cost rather than a price prediction. A real closing fee changes with the closing value. We assume dollar parity for USDT and USDC in these examples; swapping between them is a separate transaction with its own price.

Why MEXC and Gate need separate treatment

MEXC belongs in a low-fee investigation, but a single universal rate does not describe its current offers. Its June new-user announcement describes a verified-user allowance on selected contracts, followed by region-dependent prices. That is a conditional promotion, not a lifetime zero-fee account. Its API notice has a separate problem: the adjustment table and worked example disagree. We show that conflict rather than quietly choosing the cheaper number.

Gate's public fee page is dynamic and differentiates product categories and Points deductions. A fee quoted for a stock-linked perpetual is not evidence for a BTC perpetual. Its collected BTC_USDT contract response supplies default fee fields, but these do not establish the account's final VIP or Points-adjusted tariff. Gate’s contract default is shown separately from confirmed account prices.

For a manual order, the usable comparison is the row matching your account class. For automation, use the separate API price and throughput conditions below. If a promotion is volume-capped, price the intended volume above the allowance as well as the free portion. That separates three genuinely different bills instead of blending them into a single “average futures fee.”

A round-trip estimate needs two fills. On a constant 10,000 position, multiplying the maker rate by 10,000 once does not tell you what you pay to get back out. The calculator below separates the entry and exit choices. It uses the scoped schedules above, excludes VIP and token discounts, and leaves funding and execution as explicit additions. Its result is a commission estimate, not a promised total bill.

Immediate entry and immediate exit

Two taker fills produce 0 dollars in normal trading commission for the free manual profiles, 9 dollars for Hyperliquid's base profile, 10 dollars for Binance or OKX and 11 dollars for Bybit. All of those figures assume unchanged position value between entry and exit. If you close a profitable long for 10,500 rather than 10,000, the second percentage applies to 10,500. The slightly higher closing commission is caused by the value traded, not a mysterious profit tax.

Suppose the Binance and OKX trades each incur 10 dollars in commission. If one delivers an entry and exit that together are 4 dollars better, it wins the execution-and-commission comparison by 4 dollars. Ranking those two by their identical fee percentage adds no information. The comparison only becomes useful when it states what actually breaks the tie.

Passive entry followed by a market exit

Using a maker entry and taker exit changes the reference commissions to 6 dollars for Hyperliquid, 7 dollars for Binance or OKX and 7.50 dollars for Bybit. The maker saving is earned only on the quantity that really rests and fills. If half the entry fills passively and the remainder is taken immediately, calculate the two portions separately; do not assign the entire position the cheaper rate.

There is also a cost that no fee table can settle: the price may move away while you wait. Assume, purely as an example, that waiting saves 3 dollars but you eventually enter 12 dollars worse. The completed purchase is 9 dollars more expensive. That does not make passive orders a bad choice; it means a maker-only calculation should not be advertised as the inevitable price of a trade that needs immediate execution.

Frequent trading changes the importance of a small difference

For twenty unchanged 10,000 USDT round trips, two-taker commission is 200 dollars at 0.05% per side and 220 dollars at 0.055%. The 20-dollar gap is the fee advantage across 400,000 USDT of total filled value. It is not a saving on a 10,000-dollar deposit. Reusing the same collateral creates repeated trading volume and repeated commissions.

Leverage makes the debit more noticeable against margin, but the rate still applies to traded value. A 10-dollar commission is 1% of 1,000 dollars posted for a 10,000 position. It is 10% of 100 dollars posted for that same position. The fee has not increased; the collateral denominator has shrunk. This is why moving to a higher multiplier is not a fee-saving strategy.

The useful comparison is the price for the order you intend to execute. A small displayed spread at the top of the book does not establish the average price for an order that consumes several levels. Conversely, a pool-based exchange should be compared using its quoted trade output and charges, not marked as having no liquidity because it does not publish a conventional order book.

If your average entry and exit prices already measure the difference from the midpoint, they include the spread crossed and any additional movement through the book. Adding a separate spread estimate again would double-count part of the cost. An honest comparison either reports one combined execution difference or carefully decomposes it. It should not make the paid venue look worse by counting the same price disadvantage twice.

The wallet and collateral transfer can change a small-account result

A first trade on a new chain can require a withdrawal from an exchange, a bridge or a conversion into supported collateral. These are costs of reaching the venue. They should be shown separately from recurring order commissions because they are not necessarily paid on every trade. If a move costs 6 dollars and saves 1 dollar per completed trade, it needs six comparable trades just to recover that initial cost, before considering price differences.

That distinction is especially relevant for a small balance. Someone who already has usable collateral in a protocol is not in the same position as someone starting with money in a bank account. A ranking can fairly prefer a zero-commission platform for one reader and an existing centralized account for the other. The conclusion follows from their starting balances, not from changing the definition of cheap midway through the article.

For GMX or Jupiter, include the particular market's opening and closing quote, any borrowing or carry charge, and transaction execution costs. Their fee rules do not reduce to a universal maker/taker pair. We therefore keep them outside the fixed order-book tariff table rather than inventing an equivalent rate. A native quote can still beat an order-book route; it must be compared on the same exposure and holding period.

We collected settled BTC perpetual funding observations from Binance, Bybit and OKX on 4 October 2026. The shared sample ends at 16:00 UTC that day. It contains matching eight-hour timestamps, allowing one-day, three-day and seven-day comparisons. Binance timestamps with a few extra milliseconds were normalized to their settlement second. All calculations use a constant 10,000 USDT exposure, so they isolate funding rather than reconstruct a changing-position trading account.

Over the final 24 hours, a long's calculated funding cost was about 0.59 USDT on Binance, 0.85 on Bybit and 0.90 on OKX. Over seven days, it was about 8.08, 5.89 and 8.03 respectively. The lowest funding result changed with the window. An exchange that looks cheapest in a screenshot of the current rate can therefore lose over the period you actually plan to hold.

Combining the seven-day funding sum with the regular market-entry and market-exit commissions gives approximately 18.08 USDT for Binance, 16.89 for Bybit and 18.03 for OKX. Those totals exclude execution differences, conversion and any account-specific discounts. Within that defined model, Bybit beats the other two for a long despite its higher taker fee.

The short side produces a different conclusion

The net funding sums in the seven-day sample are positive, so a short receives rather than pays them. Subtracting that receipt from the same commission assumptions leaves about 1.92 USDT for Binance, 5.11 for Bybit and 1.97 for OKX. Binance and OKX are nearly tied on this partial-cost basis. The difference is too small to treat as a durable advantage once real execution is included.

This does not mean a short earns a reliable yield. The position still gains or loses as the crypto price changes, funding can reverse, and the recorded window is historical. The point is narrower: a low funding rate benefits the payer while reducing the receiver's credit. A ranking that labels lower funding as universally better is applying the wrong logic to half its readers.

What this sample can and cannot establish

The raw observations are available with the chart. They are public funding records, not our executed trades. We did not continuously rebalance a live position to maintain exactly 10,000 USDT. The fixed exposure is a comparison convention. A real BTC quantity held unchanged would acquire a different notional value as Bitcoin moved, changing each settlement amount.

We also did not substitute zero for venues lacking an equivalent collected history. Paradex and Lighter's free commission does not imply zero funding. To compare one of them with the seven-day results, obtain its settled payments for the same interval and direction. That gap is a reason to limit the conclusion to these three products, not a reason to pretend the other products performed badly.

An API order is simply an order sent by software, but exchanges can price that channel differently. MEXC explicitly separates its API tariff from web and app offers. Paradex classifies orders by submission method and interactive status. Lighter's account categories exchange different fees and timing conditions. These differences can overwhelm the commission savings advertised on a homepage.

MEXC: the documented conflict remains material

The June API update says the post-adjustment maker/taker rates are 0.06% and 0.08%. The worked example beneath it still uses 0.04% and 0.06%. Both are separate from manual promotions. On a 10,000-dollar taker round trip, those taker alternatives imply 16 or 12 dollars before funding. The disagreement is large enough that choosing one silently would change the result.

We exclude MEXC from the confirmed API price ordering until the applicable current rate is resolved by an authoritative account-specific schedule or a corrected notice. For manual trading, it remains a candidate for eligible promotions. This is more useful than either dismissing the exchange entirely or advertising a promotional manual rate to a reader who intends to automate orders.

Paradex: interactive API is an exception worth understanding

The simple claim that every API order pays Pro fees is also incomplete. Paradex documents interactive API tokens that submit Retail orders, subject to Retail limits. Exceeding those limits changes the behavior. A low-frequency interactive tool and a high-throughput automated strategy therefore do not necessarily fall into the same price category, even though both use an API.

For standard Pro flow, the published base maker/taker pair is 0.003% and 0.045%. That implies 0.30 dollars for a 10,000-dollar passive fill and 4.50 dollars for an immediate one before discounts. A maker entry and taker exit total 4.80 dollars. The zero Retail calculation cannot be applied to that Pro trade simply because both originate from the same wallet.

The decision for an automated trader should be made on its actual order pattern: the permitted channel, the account category, how often orders are cancelled and what fees appear on fills. The fastest published API allowance is not itself a price advantage, and the cheapest nominal tariff may not support the required interaction pattern.

A discount is useful only when the value saved justifies the condition required to obtain it. Suppose holding an extra token reduces your expected monthly futures commission by 12 dollars. That is the relevant fee benefit to compare with the capital committed and the token's price movement. A headline reduction of 10% sounds larger than 12 dollars, but they can describe the same offer.

Volume tiers should follow trading you intended to do anyway. Generating extra trades merely to reach a lower tier can cost more than the future saving. Compare the fees on your expected activity before and after qualification, and account for the fact that many tiers are based on a rolling window. Reaching a tier once does not necessarily lock it in forever.

Referral benefits also need to be distinguished from the standard schedule. Our buttons use the registered partner destinations, but the calculations do not assume an undocumented permanent discount. Signup credits, fee vouchers, points and withdrawable cash are different benefits. A voucher that expires before your next trade cannot be treated as cash reducing every future order.

Finally, an advertised maker rebate is a credit attached to eligible executed volume. It is not interest on idle collateral, and a resting order that never fills produces no trading rebate. Comparing negative maker rates without considering the taker exit can make an account appear to pay you for a complete trade when only one side earns a credit.

A filled entry can contain both maker and taker portions. Suppose 6,000 USDT rests at a 0.02% maker rate and the remaining 4,000 executes at 0.05% taker. Entry commission is 1.20 plus 2.00, or 3.20 USDT. An unchanged-value 10,000 taker close adds 5.00, making 8.20 in total. Applying the maker rate to the entire entry would understate that bill by 1.20 USDT.

Use filled value rather than the original requested amount when an order only partly executes. A 10,000 request that fills 4,000 creates commission on the 4,000 fill, not the unfilled 6,000. The fee schedule still applies; the quantity to which it applies has changed. The order history’s individual fills therefore explain a result that a single headline percentage cannot.

Finally, separate a reserved closing allowance from an actual closing charge. The former can reduce available margin before the trade finishes; it should not be deducted a second time after counting the executed closing fee. Reconcile commissions from fills, funding from settled events and transfers from the transfer record. This produces a bill you can compare with the estimate without counting the same deduction twice.