Skip to content

Forex Broker Execution Speed: Live Test of ASIC Brokers 2026

Written by Justin Grossbard (RG146) Fact-checked by David Levy Last updated:

In our live execution test, ACY Securities executes market orders fastest. The test ranked 21 Australian brokers on 4,018 filled orders from 2 to 13 November 2026, all live EUR/USD market orders. The ranked figure is broker time, send to fill minus the terminal’s ping to the broker’s trade server, and the field median is 66.7 ms.

Execution speed ranking of ASIC-regulated brokers

Use any column heading to sort the table, and each broker's details button opens its other figures.

All 21 brokers ranked by median broker time, from 0.01 lot EUR/USD market orders on live funded accounts, window 2 to 13 November 2026.
10 to 50 ms
1ACY SecuritiesProZero · MT5ASIC licensed27.137.331.6Better than requested 5.4%At requested price 86.8%Worse than requested 7.8%0.0%0.47

Two ACY Securities trades on 2 November 2026 closed seconds after opening because the tester's stop sizing read a wrong USD to AUD rate, fixed at 02:56 UTC. The fills and timings of those trades are valid.

2AxiPro · MT5ASIC licensed32.0142.734.4Better than requested 1.5%At requested price 97.0%Worse than requested 1.5%0.0%0.36
3Fusion MarketsZero · MT5ASIC licensed33.953.7102.1Better than requested 5.6%At requested price 89.4%Worse than requested 5.1%0.0%0.32
4FP MarketsRaw · MT5ASIC licensed35.750.339.0Better than requested 4.5%At requested price 90.9%Worse than requested 4.5%0.0%0.57
5Global PrimeECN · MT5ASIC licensed35.8206.8105.2Better than requested 5.2%At requested price 90.2%Worse than requested 4.6%0.0%0.57

The Global Prime terminal was logged in read-only until about 02:20 UTC on 2 November 2026, so it placed no orders before then. This has no effect on the figures.

6PepperstoneRazor · MT5ASIC licensed36.139.9109.0Better than requested 8.6%At requested price 80.3%Worse than requested 11.1%0.0%0.60
7IC MarketsRaw Spread · MT5ASIC licensed42.362.7112.2Better than requested 3.0%At requested price 90.4%Worse than requested 6.6%0.0%0.73
50 to 200 ms
8EightcapRaw · MT5ASIC licensed55.175.5275.0Better than requested 18.0%At requested price 72.2%Worse than requested 9.8%0.0%0.59
9Trading.comStandard · MT5ASIC licensed62.172.474.6Better than requested 1.5%At requested price 93.9%Worse than requested 4.5%0.0%1.12
10OANDAStandard · MT4ASIC licensed64.883.4138.1Better than requested 7.6%At requested price 84.8%Worse than requested 7.6%0.0%1.01
11GO MarketsGO Plus+ · MT5ASIC licensed66.777.3139.1Better than requested 10.1%At requested price 80.8%Worse than requested 9.1%0.0%0.50
12FXCMStandard · MT4ASIC licensed68.4108.1141.3Better than requested 13.6%At requested price 74.4%Worse than requested 11.9%0.0%0.87

The FXCM account was held for margin from 6 November 2026 until it was topped up, so it has fewer filled orders.

13IGIG MT4 CFD · MT4ASIC licensed87.6150.792.4Better than requested 5.6%At requested price 90.9%Worse than requested 3.5%0.0%0.76
14FOREX.comStandard · MT4ASIC licensed113.8149.1115.8Better than requested 6.6%At requested price 84.8%Worse than requested 8.6%0.0%0.86
15CMC MarketsFX Active · MT4ASIC licensed134.1167.8138.0Better than requested 7.1%At requested price 86.4%Worse than requested 6.6%0.0%not in the October 2026 capture
16VantageRAW Classic · MT5ASIC licensed134.8159.3213.0Better than requested 6.6%At requested price 83.7%Worse than requested 9.7%0.0%0.50
17Blueberry MarketsDirect · MT5ASIC licensed159.3245.7161.4Better than requested 10.1%At requested price 81.3%Worse than requested 8.6%0.0%0.62

At 22:31 UTC on 4 November 2026 four close attempts at Blueberry Markets failed with no connection, and the fifth filled seventeen seconds later. The disconnect episode stays in the figures because it was a genuine broker disconnect.

200 ms to 1 s
18Capital.comMT4 CFD · MT4ASIC licensed229.9273.1251.1Better than requested 11.1%At requested price 80.2%Worse than requested 8.6%0.0%0.74

The Capital.com account was held for margin after 2 trades on 2 November 2026 until a transfer on 4 November, so it has fewer filled orders.

19ThinkMarketsThinkZero · MT4ASIC licensed319.0394.3391.4Better than requested 14.5%At requested price 62.7%Worse than requested 22.8%0.0%0.66
20Trade NationStandard · MT4ASIC licensed345.4401.9347.6Better than requested 16.1%At requested price 71.0%Worse than requested 12.9%0.0%0.44
21AvaTradeStandard · MT5ASIC licensed354.0357.4356.5Better than requested 14.6%At requested price 71.5%Worse than requested 13.9%0.0%0.86

AvaTrade required every client to reset their password on 9 November 2026, and the test account was offline from 10:06 UTC that day until 11 November and placed no orders, so AvaTrade has fewer filled orders.

Broker time removes the terminal's network round trip to its trade server, so it shows the broker's own processing. Total time keeps the network in, and the cost column is EUR/USD all-in from the October 2026 spread capture.

Each dot is one broker. Across the chart is broker time on a log scale, with further left quicker. Up the chart is EUR/USD all-in cost from October 2026, with lower cheaper. The amber dot is rank 1. CMC Markets is not plotted because it has no October 2026 EUR/USD cost figure.

Data in this table spans 2 to 13 November 2026, with 4,018 filled orders across 21 brokers, tested from two London servers. Noam Korbl produced the results and David Levy verified them.

The ranked column is each broker’s median broker time, with the 95th percentile, the slow tail that one order in twenty reaches or exceeds, printed beside it. In my view the p95 column deserves as much of your attention as the median, because a median alone hides how uneven a broker’s fills are, and the p95 shows how far its slow orders stretch. Total time from London is printed as text and never ranked, since it depends on where you connect from.

How to cite this data: CompareForexBrokers Australia (2026). Live execution speed at 21 ASIC-regulated forex brokers, 2 to 13 November 2026. https://www.compareforexbrokers.com.au/about/execution-speed/

How the execution test works

I rank execution on live orders, because a demo account cannot show you what a broker’s server does with your money. Noam Korbl, our Co-Founder and COO, ran the test, as he runs the live spread captures. He used our own MetaTrader expert advisor, CFB-ExecTester, one copy per broker terminal, on the two London VPS servers that also carry the spread capture, with a live account at each broker’s Australian entity, funded with our own money in Australian dollars.

Every order was a real EUR/USD market order of 0.01 lot, and both the opening and the closing order were timed. Each terminal could place up to 10 round trips per UTC day, at randomised times across the trading day, each position held for a randomised 6 to 15 minutes, with a protective stop-loss order sitting at the broker on every open position. The window leaves out the weekend, Friday 20:00 UTC to Sunday 23:00 UTC, and the daily rollover, 20:45 to 22:15 UTC, and no position opened while the live spread was above three times its recent average.

The ranked figure is broker time: the time from sending the order to receiving the fill, minus the terminal’s own ping to the broker’s trade server at that moment. That subtraction strips out our London network and leaves the broker’s own processing. We rank the median broker time and print the 95th percentile beside it. It is the part of the delay your broker controls, whichever city you trade from. Total time from London, send to fill including the network, is printed as text beside the speed but never ranked, because it depends on where you connect from.

A broker needs at least 100 filled orders, open and close legs counted separately, before it can be named a winner, and a winner is named only when both ISO weeks of the window agree. I like that two-week rule, because one fast morning should not crown a winner on its own. So you can weigh speed against price, each row prints the broker’s EUR/USD all-in cost, spread plus commission, from our October 2026 spread capture; CMC Markets appears as not in the October 2026 capture, because its terminal recorded a gap from 10 to 19 October. David Levy, our Head of Content, checked every published figure against the dataset before publication. The window itself was not entirely quiet, and the incidents below are part of the record.

What the test does not tell you

The Trading Experience rating is still desk-assessed: our research desk scores it as an editorial judgement, the measured speed feeds no score or rating, and no written mapping from the test to a score has been adopted. Stop-loss and limit order fills, swap rates and a month-by-month view are not tested in this release. A fast fill is also not evidence that an account is “true ECN”, because the test does not record where an order was routed.

What happened during the test window

Three accounts lost trading time during the window: AvaTrade to the broker’s password reset, and Capital.com and FXCM to our tester’s margin rule. I would not read the shorter order counts as a warning about those brokers, because our own tester caused two of the pauses. AvaTrade made every client reset their password on 9 November 2026, which logged our account out from 10:06 UTC that day until 11 November; no orders ran in that time, trading resumed on 11 November, and AvaTrade has not given a reason. Capital.com and FXCM were held back by the tester’s margin rule, Capital.com after two trades on 2 November until a transfer landed on 4 November, and FXCM from 6 November until we topped the account up. Both brokers simply have fewer filled orders than most.

The other three incidents left the figures intact. A Global Prime terminal was logged in with a read-only password until about 02:20 UTC on 2 November, so no orders ran before then; this has no effect on the numbers. Two ACY Securities trades on 2 November closed seconds after opening, because our tester’s stop-loss sizing read a wrong US dollar to Australian dollar rate, which we fixed at 02:56 UTC that day; the fills and their timings are valid and kept. At 22:31 UTC on 4 November, four attempts to close a Blueberry Markets position failed with a “no connection” message from the broker’s server, and the fifth filled 17 seconds later; we kept it in the data as a genuine broker disconnect, one event among its filled orders.

What the results show

The fastest broker led both weeks

We name a fastest broker only when the two ISO weeks of the window agree independently, and ACY Securities cleared that bar. Its median broker time was 27.1 ms, the lowest of the 21 brokers ranked, and it led in the first week and again in the second, past the 100-order minimum a broker needs. In my view the two-week lead is what makes this worth trusting, since one strong week alone could be a quiet market doing the work. I would read a single-week leader as interesting and a two-week leader as a real signal.

Where the trade server sits

The network finding is the one I weigh heaviest. Total time from London ranks the brokers differently from broker time, because some brokers run their trade servers far from London, so the travel leg, a few milliseconds for some brokers and many times that for others, dominates the total. That is why we rank broker time and print total time as text. Your own network leg differs again if you trade from Australia, and we have not yet run a check from Sydney, so read the total-time column as our London view, not your view.

Two brokers changed between the weeks

Two brokers moved by more than 15 ms between the weeks. Axi’s median broker time was 29.8 ms in the first week and 87.4 ms in the second. Blueberry Markets’ was 147.8 ms and then 168.8 ms. Both keep their place in the table on the median of the whole window, and Axi’s wide 95th percentile comes from the same shift. We have not established the cause. In my view the shift itself is the warning for you: a two-week median can hide a change inside the window, so check the weekly numbers in this section before you treat any single ranked figure as how a broker always fills.

No broker rejected an order

Not one of the 21 brokers rejected an order in the window, and every market order we sent filled. Our test did not cover requotes, and we did not test stop or limit orders. In my view, one window with zero rejections shows these brokers handled our market orders cleanly, and it shows nothing beyond that. Because all of them scored the same, I would not use this finding to separate them.

FAQs

Which broker filled orders fastest in your live test?
Our two-week test found ACY Securities recorded the lowest median broker time at 27.1 ms, leading all 21 brokers ranked from 2 to 13 November 2026. It topped both weeks of the test, which we require before naming a winner.
What does the execution test measure?
Noam Korbl ran the test using live funded accounts at each broker's Australian entity. Each broker received 0.01 lot EUR/USD market orders, with 4,018 orders filled in total across all brokers. We rank each broker's median broker time and print the 95th percentile beside it.
Does a faster broker also cost less to trade?
No. We measure speed and cost separately, and the speed ranking ignores cost. The table prints each broker's October 2026 EUR/USD all-in cost beside its speed, so you can weigh both together.
Is execution speed tested across data centres?
No. The live test ran from our two London servers only, so it compares brokers from one location. Your own network leg depends on where you trade from, and we have not yet run a check from Sydney.

Some brokers may pay us a fixed fee or per account opened. Learn how we make money.

About Justin Grossbard

Justin Grossbard headshot

Justin Grossbard

Justin co-founded CompareForexBrokers in 2014 and has traded forex since 1998. Based in Melbourne, he leads the site's research, and has written for Forbes, Kiplinger and Finance Magnates, while the Australian Financial Review has cited CompareForexBrokers research. He holds a Bachelor of Commerce (Honours) and a Master of Marketing from Monash University. Justin is the co-founder and CEO of CompareForexBrokers.

LinkedIn · X / Twitter