Best Low-Spread Forex Brokers for 2026

Three different numbers all get called the spread. The minimum is the tightest print a broker has observed. The average is a mean taken across the whole trading day, including hours you may never trade. What you actually pay is neither: it is the spread at the specific instants your orders reach the market, and since your entries are not randomly distributed through the day, your realised average will differ from any figure a broker publishes. That gap is the single most useful thing to know about your own trading, and it is the only one of the three you can measure rather than read.

Our dataset records indicative raw EUR/USD spreads of around 0.01 pips at IC Markets, 0.1 pips at Pepperstone, 0.1 pips at FOREX.com, 0.2 pips at Interactive Brokers and 0.3 pips at FxPro. Those figures are not the cost of trading at any of them, because every one of those accounts also charges a commission. The trap is comparing a raw figure against a commission-free one: the same dataset records indicative standard-account EUR/USD spreads of about 1.1 pips at Pepperstone, 1.0 pips at FOREX.com and 1.6 pips at FxPro. Setting 0.1 against 1.1 and concluding the raw account is eleven times cheaper is arithmetic nobody should accept. Two brokers are deliberately absent from that second list: our dataset's standard-account figures for IC Markets and Interactive Brokers are not credible as commission-free retail forex pricing, so they are used here only in raw-account comparisons.

One further caveat before any comparison is worth running: a broker that is tightest on EUR/USD need not be tightest on the pair you trade. Spreads reflect the liquidity available in each individual instrument, and the ranking between firms can reorder completely between a major, a cross and an exotic. Our dataset records no spread figures for any instrument other than EUR/USD, at any broker on this page, which is a limit on what this comparison can tell you rather than a gap in the market. If you trade GBP/JPY or a Scandinavian cross, compare on your own instruments at your own hours.

The conversion is the whole exercise, and it takes one line of arithmetic. A standard lot is 100,000 units of the base currency, and on most pairs a pip is 0.0001, so a pip is worth about ten units of the quote currency per standard lot - roughly $10 on a USD-quoted pair held in a USD account. Divide the round-turn commission by ten and you have its pip equivalent, which can then be added to the raw spread to give a single all-in number. Two caveats apply before you use it: yen-quoted pairs use a 0.01 pip and must be calculated separately, and if your account is denominated in another currency the pip value is converted at prevailing rates and therefore moves.

Applied to the schedules in our dataset: Pepperstone's Razor account charges $7 per standard lot round turn on MT4 and MT5 and $6 on cTrader, which is 0.7 and 0.6 pips respectively; added to a raw spread near 0.1 pips that gives roughly 0.8 and 0.7 pips all-in, and our Pepperstone profile independently puts its Razor all-in cost at about 0.8 pips on EUR/USD. IC Markets' Raw account is recorded at $7 on MT4 and MT5 and about $6 on cTrader against a raw spread near 0.01 pips - but our dataset does not state whether those are per-side or round-turn figures, so the resulting 0.7 and 0.6 pips all-in assume round turn and need confirming with the broker before you rely on them. FOREX.com's RAW Spread account charges $7 per standard lot round turn, stated as such, against a 0.1 pip raw spread - about 0.8 pips all-in, consistent with the 0.8 to 1.1 pip range recorded in its profile. FxPro's Raw+ commission is quoted per side at $3.50 per lot, so $7 round turn, which against a 0.3 pip raw spread is about 1.0 pip all-in.

Interactive Brokers prices on a different basis and repays the effort of converting it. Its schedule is recorded as approximately 0.2 basis points of trade value with a $2 minimum per side, tiering down to 0.08 basis points at very high volume. On a standard lot of EUR/USD, 0.2 basis points of 100,000 units is about 2 units of the traded currency per side - roughly $4 round turn on a dollar-quoted pair, which is what our dataset records directly. That is about 0.4 pips, and added to a 0.2 pip raw spread gives roughly 0.6 pips all-in. Being charged on trade value rather than per lot, those dollar figures are approximations that move with the exchange rate. It lands effectively level with IC Markets Raw on cTrader at about 0.61 pips rather than clearly ahead of it - a difference of a hundredth of a pip is noise, not a ranking, and it is only visible at all once everything is expressed in the same unit.

The convention trap is worth stating separately because it silently doubles or halves every answer above. FxPro's forex commission is recorded per lot per side; Pepperstone's and FOREX.com's are recorded per standard lot round turn; Interactive Brokers' minimum is stated per side; IC Markets' is recorded with no basis given at all. Before comparing two commission figures, establish which basis each uses - and if a broker's published schedule does not say, that is a question for support rather than an assumption to make. Everything here is indicative and recorded in our dataset rather than quoted from a live schedule, and commissions change.

The rule is arithmetic rather than ideology: a raw account is cheaper only when the standard account's spread exceeds the raw spread by more than the commission's pip equivalent. Run it on the recorded figures. At Pepperstone, 1.1 pips standard against 0.1 raw plus 0.7 commission gives 0.8 - raw wins by roughly 0.3 pips. At FxPro, 1.6 against 0.3 plus 0.7 gives 1.0 - raw wins by about 0.6. At FOREX.com, 1.0 against 0.1 plus 0.7 gives 0.8 - raw wins by roughly 0.2, a narrower margin than the marketing suggests. For IC Markets and Interactive Brokers the test cannot be run at all, because our dataset holds no credible commission-free forex tier for either. Where it can be run, the answer is decided by tenths of a pip - small enough that platform choice, minimum ticket charges or a single bad fill can reverse it.

Trade size introduces a second effect that reorders the ranking entirely, and it turns on how the commission is structured. A flat per-lot commission scales with volume, so its pip equivalent is constant: $7 per lot is 0.7 pips whether you trade one lot or a tenth of one. A schedule with a minimum charge per order does not scale. On Interactive Brokers' recorded $2 minimum per side, a 0.1 lot trade - where a pip is worth about $1 - still costs about $4 round turn, which is 4 pips in pip-equivalent terms. The same schedule that produces one of the lowest all-in costs on this page at one standard lot becomes among the most expensive at a tenth of that size. Anyone trading in mini or micro lots should check for minimum ticket charges before looking at a single spread figure.

The third variable is frequency, and it decides which cost even matters. Spread and commission are charged per trade and therefore scale with turnover; overnight financing is charged per night and scales with holding period. A trader placing 200 round turns a month at one standard lot pays roughly $8 per round turn at a 0.8 pip all-in cost - about $1,600 a month, before any financing. A trader placing four round turns a month at the same size pays about $32, at which point the spread comparison you spent an afternoon on is worth less than a rounding error and the swap table is the document that matters. A low-spread ranking is a ranking for one kind of trader, and it is worth being honest about which kind you are before acting on it.

Commission is a property of the account and frequently of the platform, not of the brand - a detail that comparison tables almost never surface. Our dataset records Pepperstone's Razor commission at $7 per lot round turn on MT4 and MT5 but $6 on cTrader. IC Markets' Raw commission is recorded at $7 on MT4 and MT5 and about $6 on cTrader. In both cases the same firm, on the same instrument, prices the same trade differently depending on which terminal you placed it from.

FxPro is the sharpest case in our data and also the one to check hardest. Its forex commission is recorded at $3.50 per lot per side on MT4 and MT5, so $7 round turn, while its cTrader schedule is quoted on a volume basis as $35 per $1 million traded, which our dataset glosses as roughly $3.50 per lot round turn. If that gloss is right, the platform you pick is worth around half your commission bill at the same broker. A per-million rate is precisely the kind of figure whose meaning doubles depending on whether opening and closing are each counted as volume traded, so treat it as a question for FxPro rather than as a saving already banked.

The practical order of operations follows from that. Choose the platform on capability - whether you need MQL5 automation, a depth-of-market ladder, a specific charting workflow - and then check whether the same broker prices that platform differently before assuming the account you were offered is the cheapest one it has. Account names will not help you here: Razor, Raw, Raw+, RAW Spread, ECN and Zero are labels chosen by each firm's marketing department and mean nothing consistent across brokers. The commission schedule and the spread table are the only documents that carry information. Volume tiering is the other lever and should be read sceptically: Interactive Brokers' recorded schedule tiers from about 0.2 basis points down to 0.08 at very high volume, a genuine reduction, but our dataset does not record the threshold at which it applies, so treat it as a feature of the schedule rather than something you are likely to reach. Trading more than you intended in order to unlock a lower rate usually costs more in extra trading than the rate reduction returns.

Finally, availability constrains all of this. Our dataset records that Interactive Brokers offers neither MetaTrader 4 nor MetaTrader 5, which means a trader whose strategy depends on an MQL Expert Advisor cannot reach IBKR's pricing at all, however attractive the pip-equivalent arithmetic looks. The cheapest schedule you cannot actually use is not a cost saving.

A spread is the price of immediacy, set by whoever is willing to take the other side right now. It is tightest when many providers are quoting the same instrument simultaneously - for the majors, that means the London session and the London to New York overlap - and it widens whenever that number falls. The same mechanism explains the enormous range between instruments: EUR/USD is quoted in fractions of a pip because it is the most heavily traded pair in the world, while a thinly traded exotic is quoted in whole pips or tens of them, and no broker can source liquidity that does not exist.

The daily rollover is the most predictable widening event and the one that catches most people out. Around the platform's end-of-day, providers step back while positions are rolled to the next value date, and spreads on many instruments widen sharply for a few minutes. This is routine and expected. What is not obvious is the consequence for orders resting near the market: a spread that briefly trebles can touch a stop that the mid-price never came near. If you hold through the rollover it is worth establishing whether your broker triggers stop-loss orders and margin close-outs on the bid, the ask or the mid for each instrument, because that determines whether a widening spread alone can take you out of a position.

Scheduled economic releases produce the same effect with less predictable timing and greater magnitude, and the widening typically begins before the number prints as providers pull back ahead of it. For cost modelling this matters more than it seems: your realised average spread will exceed a broker's advertised average by whatever proportion of your entries cluster around those moments. If a large share of your trading happens in those windows, the broker's published average is simply not a description of your bill, and the only reliable figure is the one computed from your own filled trades. The execution consequences of those same seconds - rejections, fills landing beyond the requested level, stops filling through their trigger - are covered in our guide to brokers for scalping.

Exotic pairs are worth treating as a separate category rather than as expensive majors, because the cost is worse on more than one axis at once: a wider spread, larger absolute movement per unit of time, and a heavier cost to carry. Our dataset gives one clean illustration of the last of those - IC Markets' swap-free schedule, which records a holding fee of roughly $9 per lot per day on EUR/USD against roughly $101 per lot per day on USD/TRY. That is an elevenfold difference in the cost of simply keeping the position open, before a single pip of spread is paid. It is a swap-free administration schedule rather than a swap table, so it does not tell you what that broker or any other charges in ordinary financing on the pair; what it does show is how sharply the cost of holding scales away from a major.

Overnight financing derives from the interest rate differential between the two currencies in a pair, adjusted by the broker's own markup - which is why both the long and the short side of the same pair can be negative at the same time. Because spot forex settles on a T+2 basis, the weekend is normally collected in a single triple-rate day, commonly Wednesday for currency pairs, though the day can differ by asset class and is set per instrument by the broker. It is a published property of each symbol rather than something to infer, and our guide to MT5 brokers covers where to read it in the platform.

The scale relative to a tight spread is what makes this section belong on a low-spread page at all. At an all-in cost of about 0.8 pips, a round turn on one standard lot costs roughly $8 - a one-off charge. Financing is a subscription: it recurs every night the position is open, and on a trade held for several weeks it routinely exceeds everything paid in spread and commission combined. This is the single strongest reason not to read a low-spread ranking as a general broker ranking. Our dataset records swap-free administration fees but not the underlying swap rates for any broker on this page, so the actual numbers have to be read from each firm's own swap table for the specific pair and direction you intend to hold - and if you hold for weeks, that table decides more of your outcome than everything else on this page put together.

Swap-free and Islamic accounts do not remove the cost of holding; they change its form, and the replacement is often more expensive for a long hold than the swap it replaces. Our dataset records Pepperstone offering swap-free on request to residents of eligible countries, with FX and precious metals positions held beyond five days incurring an administration fee of $100 per standard lot. IC Markets replaces swaps with holding fees charged every calendar day including weekends, with a five-day grace period on most instruments, two days on USD/JPY and GBP/JPY, and no grace period at all on energy CFDs, where fees begin from day one. FxPro allows indices, futures and shares to be held indefinitely swap-free but applies administration fees to forex, metals and crypto after a grace period our data records as unpublished, which is itself a finding. For FOREX.com and Interactive Brokers our dataset records no swap-free account; FOREX.com's own note adds that none is listed and tells applicants to confirm with support if they need one.

One structural detail in that list deserves separating out, because it directly contradicts the premise of this page. Pepperstone's swap-free terms are recorded as being based on Standard account conditions with 1.0 to 1.2 pip spreads. A trader who needs swap-free status may therefore be paying a wider spread and an administration fee simultaneously - the opposite of the raw-plus-commission structure the rest of this page is about. Our data also records a $200 minimum deposit for that account against a nominal $10 elsewhere. If swap-free is a requirement for you, model the total on the account terms actually attached to it rather than on the raw account's headline pricing.

Currency conversion is the most reliably overlooked charge because it is never a line item you can point at, and it arises in two separate places. At the edges of the account, a bank denominated in one currency and a trading account in another means paying a conversion on the way in and again on the way out. Inside the account, trading an instrument that settles in a currency other than your base currency means realised profit and loss, commission and financing are each converted at the broker's own rate rather than the market's. The fix is structural rather than vigilant: choose a base currency that matches both your bank and the instruments you actually trade, and most of the exposure disappears without you having to monitor anything.

Then there are the fixed charges. Our dataset records FxPro applying an inactivity fee after six months - $15 as a one-off, then $5 a month - while Pepperstone's profile records no deposit or withdrawal fees. For the other three brokers on this page our data records no deposit, withdrawal or conversion charges either way, which means unrecorded rather than absent. Minimum deposits vary sharply: $0 at Interactive Brokers, $0 at IC Markets' global entity with around $200 suggested for institutional-grade pricing, a nominal $10 at Pepperstone in most regions, $100 at FxPro which recommends $1,000 for comfortable trading, and $100 at FOREX.com which recommends $2,500 for meaningful flexibility. A minimum deposit is not itself a cost, but a recommended balance is a useful hint about the account size at which a given pricing structure is designed to make sense.

The largest uncosted item is not a fee at all. Slippage is a real cost, and so is an execution model that fills you late or rejects you in fast conditions - a broker quoting 0.1 pips tighter whose fills routinely land half a pip away is more expensive, not less, and no spread table will ever show it. That is the subject of our brokers for scalping guide, and it belongs in any honest cost comparison rather than being treated as a separate topic. Usability carries a similar hidden price: Interactive Brokers' profile pairs one of the lowest recorded all-in costs on this page with a steep platform learning curve, no MetaTrader support, no swap-free option and customer support noted as slow. Those are genuine trade-offs against a fraction of a pip, not footnotes.

The model is short enough to write on one line, and every input comes from your own trade history and the broker's published schedule. Annual cost equals round turns per year multiplied by lots per trade multiplied by all-in cost per lot, plus nights held multiplied by lots multiplied by the swap per lot per night, plus fixed charges such as inactivity, conversion and withdrawal fees. Take the all-in cost per lot from the pip-equivalent arithmetic above, the swap per lot per night from the broker's own swap table for your specific pair and direction, and the trade count and average holding period from your existing statements rather than from what you intend to do.

The reason to build it is that the ranking inverts depending on the inputs. Consider two traders using the same broker and the same one-lot size. The first places 40 round turns a month and is flat by the close: at 0.8 pips all-in, roughly $8 per round turn, that is about $320 a month with essentially no financing, and a tenth of a pip difference between brokers is worth about $40 a month to them. The second places four round turns a month and holds each for an average of 15 nights: their transaction cost is about $32 a month, but they accumulate 60 lot-nights of financing, which at any realistic swap rate dominates their entire bill. The broker that is cheapest for the first trader can be the most expensive for the second, and no comparison table can resolve that for you because it depends on numbers only you have.

Two cautions on using the result. The first is a direction-of-causation problem: a lower cost per trade makes higher frequency look affordable, and trade count is the input traders control least well once they have started. A cheaper account that quietly triples your turnover is not a saving. The second is that a cost model is a model of one component. It does not capture slippage, it does not capture the trades you took because a platform made them easy, and it does not capture the cost of being on a platform you find hard to use under pressure.

Whatever the model says, be clear about what cost control does. It improves an expectancy; it does not create one. Most retail accounts lose money trading leveraged forex and CFDs, and every firm authorised in the UK and the EU is required to publish the percentage of its own retail client accounts that do, recalculated quarterly over the preceding twelve months with all costs and charges included. Read that figure for the specific entity you would open with, next to the commission schedule. It is one of the few genuinely comparable numbers a broker is obliged to show you.

Design the pricing test so it answers the question you actually have, which is not what a broker's spread can be but what yours turns out to be. Sampling at fixed intervals across the hours you genuinely trade, including at rollover and through at least one scheduled release, gives you the shape of the distribution rather than its best point - and the whole exercise collapses unless competing brokers are sampled at the same instant, since a few minutes is enough to invert the result. What no pre-funding test can give you is the spread you personally pay, because that depends on when your orders arrive. Once you are live, compute the realised spread on your own filled trades and set it beside the average the broker advertises. The gap between those two numbers is the most useful figure in this entire comparison, and it is the one nobody publishes for you.

Verify the entity as carefully as the price, because pricing, leverage and the protections that apply are all entity-specific rather than brand-specific. Our dataset records independently checkable licences you can look up on the regulators' own registers: Pepperstone under FCA FRN 684312, CySEC 388/20, ASIC AFSL 414530 and SCB Bahamas SIA-F217; IC Markets under ASIC AFSL 335692 and CySEC 362/18; FxPro under FCA FRN 509956 and CySEC licence 078/07; FOREX.com through Gain Capital Group LLC at NFA ID 0339826 and StoneX Financial Ltd at FCA FRN 446717; and Interactive Brokers under FCA FRN 208159, FINRA CRD 36418, ASIC AFSL 453554 and MAS CMS100917. Check them on the regulator's site, never on the broker's own page, and confirm which of a group's entities would actually hold your account. Retail leverage at the tier-1 entities above is recorded at 1:30 on major pairs under FCA, ASIC, CySEC and BaFin rules, and 1:50 for US retail forex majors; the higher figures some of these groups advertise belong to offshore entities carrying different protections.

Cheap is not the same as clean, and a page ranking brokers on cost has an obligation to say so. Our profiles record an Australian class action filed in 2024 against IC Markets alleging misleading conduct in the supply of CFDs to retail clients, and an ASIC finding in November 2023 that Pepperstone had breached leverage limits, after which it compensated more than 1,500 clients. Neither is on its own a reason to rule a firm out - both are matters of public record at firms that remain tier-1 regulated - but both belong in the weighing, and a fraction of a pip is a poor reason to skip that step.

Everything on this page is editorial opinion scored against our published methodology, not a statement of fact or a recommendation, and every figure quoted is indicative and subject to change by the broker at any time. Trading leveraged forex and CFDs carries a high risk of losing money rapidly. Lower costs improve the arithmetic of a strategy that works; they do nothing for one that does not.