TL;DR: If your handbook says 91 is fine, the premium grades don’t pay for themselves. In Brisbane right now 95 runs 17.5c/L over 91 and 98 runs 27.0c/L over — and even after crediting a 2% economy gain, 95 still costs about $1.06 more per 100km and 98 about $1.81 more. Over 12,000km a year that’s roughly $127 wasted on 95, $217 on 98. The premium is a near-fixed markup (~15c and ~23c across 2.5 years), not a scam that balloons at the peak — but on a 91-tuned engine it’s still money for nothing. If your manual requires 95 or 98, ignore all of this and pay the gap: that’s where premium earns its keep.
The question
You’re at the bowser, your car runs 91, and the 95 and 98 buttons are right there promising a cleaner engine and better economy. The station’s signage leans on it. So the money question: on a car that only needs 91, does the premium ever pay for itself, or is it a tax on worry?
I did it the only way that settles it — cents-per-100km and dollars-per-tank, on Brisbane numbers, both today’s and the long run.
What I checked
Every price here is the Queensland Government feed — the board price each station reported, not a driver’s guess at the pump. Same feed that fills the map.
Two cuts. Live, 30 July 2026: median ULP 91 is 196.9c/L (439 stations), 95 is 214.4c/L (304 stations), 98 is 223.9c/L (441 stations). That’s a +17.5c/L premium for 95 and +27.0c/L for 98.
(A quick aside on the counts, since 98 shows more stations than 91: that’s real, not a slip — 98 is stocked almost everywhere 91 is, and a handful of premium-only or 98-first sites tip the count over.)
Then the long run, because a single day can flatter or damn either grade. I built a weekly panel of every Brisbane-metro station from January 2024 to July 2026 — 126 weeks, about 431 stations reporting 91 — took the median of each grade each week, and averaged the gap. Across 2.5 years the premium is +15.3c/L for 95 (median 15.0, range 12–27) and +23.1c/L for 98 (median 23.0, range 20–30). One gap I have to flag: the Mar 1–Apr 27 2026 stretch went quiet in the feed, so I cut it out of every figure rather than carry a fake price forward across it. The station counts wobble a little post to post — 439 reporting 91 live today, ~431 across the panel, fewer for 95 (304 live, ~312 across the panel) — because feed coverage drifts week to week; none of the medians move on it.
A note on method, because Karen measures this differently and I don’t want the two reads to look like they’re fighting. She goes pump-by-pump — same station, same day, then medians the gaps — because 95 sells at a slightly cheaper set of stations, so a straight citywide median-of-medians (which is what I do here) can flatter the gap a touch. The reassuring part: her pump-by-pump number and my citywide one land in the same place — about 15c on 95, about 23c on 98 over the long run, ~17c and ~26–27c right now. So the citywide read isn’t inflating this. And if it’s off at all, it’s off upward — meaning the true same-pump gap is if anything a shade smaller, which only makes the case for skipping premium stronger, not weaker.
Cents-per-litre is the wrong unit, though. The whole case for premium is that it goes further, so the fair test is cents-per-100km after crediting the economy gain. The catch: on an engine the manufacturer says can run 91, that gain is small — published guidance puts it at roughly 1–3%. I used 2% for the headline. That number is external to our price data, so I’ll flag it hard in the limits; the c/L, per-100km and per-tank figures are all straight from the feed.
What the data shows
First, kill the myth that premium is a peak-time rip. If the majors were using 95/98 to gouge, the premium would balloon when 91 spikes. It doesn’t. Across the cycle the correlation between the premium and the 91 price level is weakly negative (−0.31 for 95, −0.36 for 98) — bin the 2.5 years by 91 price and the gap runs 15.4c at the cheapest fifth and 14.1c at the dearest fifth for 95, 23.5c to 22.1c for 98. So it’s a near-fixed markup that, if anything, narrows slightly when 91 is at the top of the cycle. That’s the tell for today’s live 17.5c and 27.0c: they sit above the 2.5-year averages of 15.3c and 23.1c, and it is not a peak effect — near the peak the gap should be narrower, not wider. What’s actually happening is the premium has drifted up right through 2026: since May the weekly gap has run 17–19c on 95 and 25–27c on 98, well above the long-run mean, whatever the cycle is doing. Right now it sits at about 17c on 95 and 26c on 98 — the same band Karen quotes for mid-2026 (closer to 17c, nearer 26c), so the two reads line up rather than leaving you to guess. It’s a recent widening, not a peak spike. Either way, the direction of the answer doesn’t change at the trough.
Now the money, at 8 L/100km on today’s live premiums, after the 2% economy credit:
Before the economy credit, 95 adds $1.40 per 100km and 98 adds $2.16. The 2% gain claws a little back — down to +$1.06 and +$1.81 — but not nearly enough to break even. For 95 to actually pay for itself at today’s premium you’d need an 8.9% economy gain; for 98 you’d need 13.7%. A 91-tuned engine realises 1–3%. The maths never closes.
The verdict
On a car the manufacturer says can run 91, neither premium grade pays for itself in Brisbane — not today, and not on the 2.5-year average. 95 costs you about $1.06 more per 100km after the economy credit, 98 about $1.81. Per 50L tank that’s $8.75 more for 95 and $13.50 for 98 at the pump; per year at 12,000km, roughly $127 wasted on 95 and $217 on 98 (on a 50L/12,000km basis — Karen runs a bigger 60L/20,000km car, so her dollar figures are larger; the c/L gaps underneath both are the same).
The one hard exception, and it’s not a small one: if your handbook says 95 or 98 is required — many turbo, European and performance engines do — that premium isn’t optional and this whole post doesn’t apply to you. Use what the manual demands; there the higher grade earns its keep and running 91 costs you power and can do harm. The finding here is only for engines the maker says will run 91.
So what for you
Check your handbook for one word: recommended or required. If it’s recommended (or it just says 91), the premium buttons are a habit worth dropping — put the ~$127-to-$217 a year back in your pocket and run 91. The knock sensor pulls timing back a hair; you lose a sliver of pull and economy, no damage.
And whichever grade you settle on, the saving that dwarfs the octane question is which servo — the 91 spread across Brisbane hit 38.7c/L between the cheapest and dearest postcode on one day earlier this month (the cheapest-postcode read), more than double the whole 17.5c 91-to-95 gap. Picking the right servo beats picking the octane, every time. You can see today’s cheapest near you on Pinly Fuel, straight off the Queensland Government feed, so the number on the map is the number at the pump.
The honest limit
The 1–3% economy gain on higher octane for a 91-recommended engine — I used 2% — is external published guidance, not our data; it’s the biggest lever in the sum, so plug in your own real figure before you trust the verdict. The 8 L/100km, 50L tank and 12,000km/year are illustrative — scale them to your car; the c/L premiums are exact regardless. These are Brisbane-metro medians (petrol, within 30km of the CBD); 95 has thinner coverage (304 stations live, ~312 historically) than 91 or 98, so it carries a touch less certainty. Today (30 Jul) the live 17.5c/27.0c premiums run above the 2.5-year averages of 15.3c/23.1c — but that’s a 2026 drift, not a cycle-peak effect (near the peak the gap actually narrows a touch, and the argument holds at the trough too, where the long-run gaps stay near 15c and 23c). Scale it either way and premium still doesn’t pay on a 91-tuned engine. Premiums are median-of-medians; a minority of premium-heavy or discount sites sit well off it. And none of this is engineering advice — if your manual requires premium, that trumps every number here.
— Nate