Reviews  ·  Feature Drive  ·  The Rotary File

First Flight

The Mazda Cosmo Sport Series I

343 cars, one impossible engine, and the ten months Mazda bet everything — then took it to the hardest race on Earth to prove the point.

Carveyance  /  Mazda Cosmo Sport 110S  /  2026

There are exactly 343 of them. Not 343,000 — 343. Built by hand, over ten months, in a factory that six years earlier hadn't yet figured out how to make its own engine spin without eating itself alive. The Mazda Cosmo Sport Series I is one of those rare machines where the story is, if anything, more thrilling than the specification sheet — and the specification sheet is plenty thrilling on its own.

This is the car that had to work. Not sell well, not win awards, not even necessarily make money. It had to work, once, in public, or Mazda's entire twenty-year rotary gamble was dead before it started. It worked. And driving one today still feels like sitting inside the moment a company found out it was going to survive.

The Setup

An Engine That Wasn't Supposed to Be Possible

By the early 1960s, Japan's Ministry of International Trade and Industry was quietly pressuring the country's smaller automakers to merge into the giants. Mazda — still going by its old name, Toyo Kogyo — needed a reason to stay independent that Toyota and Nissan couldn't simply copy. In 1961 it licensed a strange, largely unproven idea from Germany: the Wankel rotary engine, a design with no pistons, no valves, and, as it turned out, no shortage of ways to fail.

Mazda handed the problem to a specially assembled team of 47 engineers who became known, half-jokingly and then completely sincerely, as the “47 Ronin,” after the samurai of Japanese legend. Their leader, Kenichi Yamamoto, had returned to a Hiroshima flattened by the atomic bomb as a young man; for him this was never only an engineering brief. He told his team as much, in words that have followed the rotary ever since.

For six years, the rotary fought back. Chatter marks — worn grooves the engineers grimly nicknamed “the devil's fingernail marks” — kept scoring the rotor housings, appearing after as little as 200 hours of running. The tips of the spinning rotor were vibrating against the housing wall and gouging it. No other manufacturer chasing the Wankel, including NSU, the Germans who invented it, ever fully solved it. Mazda did — with reworked apex seals (they drilled tiny cross-holes to kill the vibration), a carbon-and-aluminum sealing material, and a deliberate drop of oil metered into the fuel to keep everything alive at speed. Then they built a car around it, ran it past 700,000 test kilometers at their new Miyoshi proving ground, and on May 30, 1967, they put it on sale.

That car was the Cosmo Sport, and the first 343 of them — the Series I — are the purest version of the whole bet, before Mazda had even had time to second-guess a single decision.

Design

Space Age, By Way of Turin and Detroit

Mazda named it for the space race — Cosmo, as in cosmos, as in the era's total obsession with rockets and orbit — and the shape lives up to it. A long, low nose. A glasshouse sanded down to the minimum possible surface area. A fastback tail with round lamps set into it like afterburners. It's not a copy of anything, but it isn't shy about its influences either: Mazda's in-house stylist, Heiji Kobayashi, borrowed the swept, jet-age roofline from — of all things — the 1961 Ford Thunderbird, and leaned on Pininfarina's Ferrari 400 Superamerica for the nose and proportions.

If you look at it and think Lotus Elan, you're not imagining things — it's speaking the exact low-and-long European GT vocabulary the Elan and the Europa were speaking that same moment. But the documented influences on record are Italian and American, not British. It's a car that looks expensive, looks fast standing still, and looks like nothing else built in Japan in 1967 — because nothing else built in Japan in 1967 was trying to do what this was.

The Powerplant

982cc, Two Rotors, and a Sound Nothing Else Made

Pop the featherweight hood and there's almost nothing there — which is the entire point. The 0810 twin-rotor engine displaces just 982cc (two 491cc chambers) and makes 110 PS at a screaming 7,000 rpm, with 130 N·m of torque arriving low, at 3,500 rpm, and staying remarkably flat all the way to a 9,000 rpm redline that a contemporary four-cylinder simply could not survive attempting. There are no reciprocating pistons slamming up and down fighting their own inertia — just two triangular rotors spinning smoothly inside their housings, twin spark plugs per chamber lighting the mixture through dual distributors, a Hitachi four-barrel carburetor feeding it, and that oil-injection system quietly doing the job Mazda's engineers spent six years learning to trust.

The result is an engine that doesn't rev so much as it spins — smoother the harder you push it, with none of the buzz, hesitation, or vibration period four-bangers dealt out as a matter of course. And it sounds like nothing else on the road in 1967: not a growl, not a bark, but a rising, turbine-like wail that climbs and climbs and never seems to run out of breath. The men who'd wrestled it into being knew exactly what they had. So did the first journalists brave enough to write about it.

That was the thing that unsettled people: you could hold it at 7,000 rpm all afternoon and it would drop back to idle as unbothered as a kettle going off the boil. Nothing with pistons behaved like that.

The Numbers

Series I, By the Sheet

Behind the Wheel

Every Bit as Thrilling as the Legend Says

Drop into the Series I's low bucket seat and the first thing you notice is how little car there is around you — barely 940 kilograms, a wheelbase shorter than what Mazda would give the Series II just months later, a big wooden wheel in your hands, and a cabin that puts you close enough to the road to feel every texture of it. Turn the key and the rotary doesn't shudder to life so much as simply spin up, instantly, with a smoothness that genuinely startled period road testers used to the shake of a cold four-cylinder.

Off idle there's a flat, useful wall of torque that makes the car feel far bigger than 982cc has any right to — this is not a peaky, gutless small-bore engine you have to flog to get moving. But the real trick happens above 4,000 rpm, where most 1960s engines are already begging for mercy and the Cosmo's rotary is just clearing its throat. Push it toward that 9,000 rpm ceiling and the car doesn't get harsh, it gets eager — smoother, if anything, the harder you lean on it, with a rising howl that period reviewers described as genuinely addictive.

The steering is quick and light, unassisted in the way that actually rewards a driver rather than punishing one, and with so little weight over the nose the Cosmo darts into a corner with an eagerness that outran almost anything else wearing a Japanese badge that year. Body control is tidy, roll is minimal, and the whole car has that specific, disappearing-into-the-seat lightness that only genuinely small, genuinely light sports cars ever manage.

It is, in short, a joy — right up until you find its one real limit. The unassisted brakes need a firm, deliberate shove rather than a light dab, and the short-wheelbase chassis, magnificent through a tight second-gear corner, gets noticeably busier and less settled the longer you hold a genuinely high speed. That's a real, documented shortcoming — and one Mazda itself would address within the year by stretching the Series II's wheelbase by 150mm. None of it erases the experience. It just means the Series I rewards a driver who respects it, rather than one who simply floors it and hopes.

Where It Belonged

The Backroad, Not the Autobahn

This is not, and was never built to be, a car for holding triple-digit speeds down a featureless European motorway. That's precisely the terrain where its short wheelbase gets nervous and its unassisted brakes start asking questions you don't want to answer at speed. Put it somewhere with a rhythm instead — a mountain pass, a coastal two-lane, anywhere with second- and third-gear corners strung together with a bit of elevation change — and the Cosmo Sport comes fully alive.

Light weight rewards quick direction changes. The rotary's flat torque curve means you're rarely caught in the wrong gear. And that 9,000-rpm top end gives you a genuine reason to hold a gear right to its limit, chasing the next apex. In town it's a different pleasure entirely: that turbine-smooth idle and flat low-end torque make it wonderfully civil in traffic — more so than any comparable four-cylinder sports car of 1967, which is exactly the everyday usability Mazda needed a flagship to demonstrate. The backroad is where the Series I stops compromising and simply shows off.

The Proving Ground

Eighty-Four Hours to Convince the World

Here is the part the specification sheet can't tell you, and the part that turned a clever engine into a legend. Reviews were rapturous, but the public wasn't buying the rotary's story — NSU had beaten Mazda to market with a rotary and it had a reputation for fragility, so the whole idea was tarred as delicate. Praise in a magazine wasn't going to fix that. Mazda needed to prove the engine could survive, in public, in front of people who were sure it couldn't.

So it picked the single most brutal way imaginable. On October 20, 1968, two Cosmos lined up for the Marathon de la Route — the 84-hour endurance race at the Nürburgring, three and a half continuous days around the full 28-kilometer circuit, a race so savage it's where men like Jacky Ickx and Vic Elford learned their trade. There were easier ways to prove a point. Mazda chose the hardest one on Earth, precisely because the point wouldn't stick otherwise. The engines were deliberately detuned and capped at 130 PS for durability, fitted with a clever intake that switched from a side port to a peripheral port as revs climbed, and sent out into a 58-car field to simply survive.

They didn't just survive. For most of three and a half days the two little Cosmos ran fourth and fifth, hunting far more established machinery. Then, in the 82nd hour — with a finish almost in sight — the all-Japanese car broke a rear axle and pitched off the course into the Eifel forest. The other, crewed by three Belgians, kept going and took the flag fourth overall, beaten only by Porsche and Lancia. An unknown Japanese company, an engine the world thought was made of glass, an 84-hour race at the Nürburgring — fourth. It remains the Cosmo's only competition outing. It didn't need a second.

The Ledger

What It Got Right — and Where It Fell Short

What it got gloriously right

The rotary is the headline, and it earns it: a small, light, impossibly smooth engine that revs like nothing else built that decade, delivers real low-end torque despite its tiny displacement, and does it all with a sound and a character no piston engine of the era could fake. Add a genuinely striking, one-of-a-kind design, steering and agility that punched well above its country's reputation in 1967, and the simple, undeniable fact of being first — the first mass-producible car anywhere built around a twin-rotor engine — and you have a car that was never going to be forgotten, no matter how few were built.

Where it came up short

The honesty owed to a car this good: the unassisted brakes are genuinely under-specified for the performance on tap, and the short wheelbase turns nervous exactly when you'd want it most composed — at sustained high speed. Mazda's own engineers weren't fully satisfied either, which is precisely why the eccentric-shaft bearings, the wheelbase, the gearbox, and the brakes were all revised for the Series II less than a year later. Fuel economy, while reasonable on paper, reflected the rotary's inherent lower thermal efficiency — a trait that would shadow every Mazda rotary that followed, right through to the RX-8. And practically speaking, almost nobody could actually buy one: 343 hand-built cars, sold overwhelmingly within Japan, meant the Series I was a unicorn on the day it launched.

The Frame

Failure, or Just Never Meant to Sell?

It's tempting to look at 343 units and call the Series I a commercial failure. That's the wrong frame entirely. The Cosmo Sport was never given the tools to be a mass-market success — no left-hand-drive version was ever built, which meant the largest car market in the world, the United States, was never even a possibility, and beyond a handful of cars quietly re-badged for other right-hand-drive markets, it essentially never left Japan. That wasn't a marketing failure. It was never the assignment.

Mazda hand-built roughly 30 cars a month specifically so its engineers could gather real ownership data before committing to a bigger run — which is exactly what happened, with the wheelbase, brakes, gearbox, and bearings all revised for the Series II that replaced it in July 1968, barely ten months after launch. Judged as a mass-market product, the Series I looks like a footnote. Judged as what it actually was — a rolling proof-of-concept built to convince the world, and Mazda's own dealers, that the rotary was real and reliable — it did its job completely. It's difficult to call a car a failure when the company built around its engine is still telling this exact story six decades later.

The Field

What It Was Up Against

In its own country, the Series I's only real rival for outright halo-car status was the Toyota 2000GT — built with Yamaha, gorgeous, and priced at roughly $7,150 in the U.S. (where it, too, was barely sold), against the Cosmo's considerably more attainable ~$4,100. Toyota's flagship out-prestiged it; Mazda's undercut it by nearly half while offering an engine nobody else on Earth could put in a showroom car. Further down the domestic ladder sat the smaller Datsun Fairlady 2000 and the high-revving little Honda S800, neither remotely in the same performance class.

Internationally — even though the Cosmo was never sold there — it lived in the same conceptual space as the Jaguar E-Type, the Porsche 911, the Alfa Romeo Duetto Spider, and, yes, the Lotus Elan: small-displacement, lightweight, driver-focused GTs built by companies willing to bet everything on one engineering idea. The Cosmo Sport is the only one of that entire group built around an engine literally nobody else in the world had ever successfully sold to the public. And at the Nürburgring, it beat the road cars' racing cousins to the line.

What It Led To

The Line That Never Stopped Spinning

The Series I's run ended, by design, in July 1968, replaced by the Series II — a longer-wheelbase, more powerful (128 PS), five-speed, power-braked evolution of the same idea, built in far greater numbers (roughly 1,176 units) through 1972. Everything the Series I taught Mazda went directly into that car, and from there into the Cosmo nameplate's later generations, and ultimately into the RX-2, RX-3, RX-7, RX-8 — and, two decades on, into the four-rotor 787B, the first Japanese car to win the 24 Hours of Le Mans outright and still the only winner without pistons. The man who led the 47 Ronin, Kenichi Yamamoto, went on to run the whole company.

The original 343 didn't get discontinued so much as graduate — Mazda's own field-testing program, wearing license plates, out in the real world, doing exactly the job the company needed done before it scaled up.

The Verdict

Three Hundred and Forty-Three

Three hundred and forty-three cars is not a production run. It's a proof. Every Cosmo Sport Series I that survives today is a physical copy of the moment Mazda proved, to itself as much as to anyone else, that the impossible engine actually worked — smooth, fast, gloriously strange, and completely unlike anything else on the road. It wasn't built to sell in volume, and it didn't. It was built to make a company's survival case in sheet metal and spinning triangles — and then it drove eighty-four hours around the Nürburgring to underline the point in front of everyone who doubted it.

Sixty years later, that case still holds up. On a good back road, in third gear, right around 8,000 rpm, it holds up completely.