Sa Calobra, measured: the 10 km you have to climb to leave
Build a week around this climb →
Sa Calobra is the only climb on this island you cannot choose to skip once you have started it. The Ma-2141 runs from the Serra’s spine down to a cove on the north coast and stops there. There is no onward road. You descend 10.0 km to the sea, and then you climb the same 10.0 km back out, or you wait for a boat.
- Length
- 10.0 km
- Height gained
- 708 m
- Average gradient
- 7.1%
- Steepest kilometre
- 8.8%
from 15 m at the port to the top of the road
summit minus the lowest point on the road; 2 of 75 published climb-sides
the one figure every source agrees on
from km 8.2
The climb, kilometre by kilometre
What makes Sa Calobra unusual is not any single ramp. It is that there is barely a metre of relief anywhere on it. Antonio Parietti laid the road out to hold a near-constant gradient, and the measurement shows he succeeded: of its 10.0 kilometres, the gentlest full kilometre and the steepest are 5.9% and 8.1%. On most climbs of this length that spread is three or four times wider.
Every kilometre of the Ma-2141, climbing from the port
| Kilometre | Gradient | Altitude at the end |
|---|---|---|
| 0.0 – 1.0 | 7.2% | 87 m |
| 1.0 – 2.0 | 5.9% | 146 m |
| 2.0 – 3.0 | 7.0% | 216 m |
| 3.0 – 4.0 | 7.2% | 288 m |
| 4.0 – 5.0 | 7.2% | 360 m |
| 5.0 – 6.0 | 6.6% | 426 m |
| 6.0 – 7.0 | 6.3% | 488 m |
| 7.0 – 8.0 | 7.2% | 561 m |
| 8.0 – 9.0 | 8.1% | 642 m |
| 9.0 – 10.0 | 8.0% | 723 m |
How we got this Elevation sampled every 10 m from the Instituto Geográfico Nacional's 5 m national terrain model along the OpenStreetMap centreline of the Ma-2141, then smoothed over 50 m. The final row is short where the climb does not end on a whole kilometre.Computed from IGN Elevacion4258_5 (CC BY 4.0 scne.es) and OpenStreetMap (ODbL).Computed 5 August 2026.
How much of the climb sits at each steepness
| Gradient | Distance | Share of the climb |
|---|---|---|
| downhill | 0.1 km | 1% |
| 0–4% | 0.7 km | 7% |
| 4–6% | 2.6 km | 26% |
| 6–8% | 3.4 km | 34% |
| 8–10% | 2.5 km | 25% |
| 10%+ | 0.7 km | 7% |
How we got this Measured over 100 m windows — short enough to be meaningful, long enough that a 5 m terrain model's own vertical noise does not dominate.Computed from the same 10 m elevation series.Computed 5 August 2026.
The 682 m everybody publishes turns out to be the consensus of one language
The English-language sources agree with each other and with almost nothing else. Wikipedia, PJAMM and OpenStreetMap’s own tag on the pass node all put the top of this climb at 682 metres. When we sampled Spain’s national terrain model at that exact node it returned 723 m, and three further elevation models sided with us rather than with them.
That looks like us against everybody, and it is not. quaeldich.de — the climb database a German rider arrives with — heads this pass «Coll dels Reis (728 m)», five metres above our own reading and forty-six above theirs. It is not the only figure on which it stands where we stand.
The whole headline set, as each publisher states it
| Published by | Publishes in | Length | Height gained | Gradient | Summit |
|---|---|---|---|---|---|
| quaeldich.de | German | 10.1 km | 710 m | 7.0% | 728 m |
| PJAMM Cycling | English | 10.1 km | 654 m | 6.3% | 682 m |
| passzwang.net | German | 10.0 km | 650 m | “knapp 7%” | 682 m |
| Our measurement (IGN 5 m) | — | 10.0 km | 708 m | 7.1% | 723 m |
PJAMM's metric figures are converted from 6.3 mi and 2,146 ft; that source publishes no metric set of its own. checked 5 August 2026.
Four quantities, and quaeldich lands inside one reporting unit on all four: +0.1 km on length, +2 m on height gained, −0.1 points on gradient and +5 m on the summit. Of every source this site holds for this climb, the closest published agreement with our own terrain model is the German one, and no English-language source is near it.
How we got this Each source’s four figures differenced against ours, and the differences added after scaling each to its own reporting resolution — 0.1 km, 10 m, 0.1 percentage points, 10 m — so that no single quantity decides the ranking on units alone.Computed from The length, height gain, average gradient and summit altitude each publisher states, against our own measured record for the same climb.Computed 7 August 2026.
Two measurements can agree and both be wrong, so this is not proof that we are right. What it does change is the shape of the disagreement: 682 m is not the number everybody else publishes, it is the number one language publishes. Look this climb up in German and the higher figure is the first one you meet.
The altitude at the top of the Ma-2141, according to each source
| Altitude | Published by | In their words |
|---|---|---|
| 682 m | Wikipedia (Sa Calobra) | Given as “the pass on 682 metres (2,238 ft) altitude”. |
| 682 m | PJAMM Cycling | Their climb page ends the ascent at “Coll dels Reis at 682 meters”, over 6.3 mi with 2,146 ft gained at 6.3%. |
| 682 m | OpenStreetMap node 1123226364 | The elevation tag on the pass node itself. |
| 682 m | passzwang.net | Takes the figure off the sign on the pass — «laut Passschild» — and notes in the same breath that altitudes «vielfach über 700m» circulate for it. A German source that was already recording this disagreement before we were. |
| 728 m | quaeldich.de | The German-language climb database heads the pass «Coll dels Reis (728 m)». It is 5 m above our own reading and 46 m above the figure the English-language sources share, and its length, height gain and gradient for the climb are within one unit of ours as well. The same publisher’s page for the Coll de Sa Bataia puts the pass 72 m higher again — «Coll dels Reis (mit 800 m der zweithöchster Pass)» — so 728 m is the figure its own pass page states, not a figure it states consistently. |
| 723 m | Our measurement (IGN 5 m national terrain model) | EU-DEM 25 m reads 724.4 m, SRTM 30 m 725.0 m and Mapzen 725.0 m at the same point. |
The split is not ours against everyone else’s, and it stopped being that the day the German corpus was read: the English-language sources share 682 m, and the German climb database publishes 728 m. Ours sits between them, 5 m from the higher one. The gradient everyone quotes — 7.1% — is the one figure every source here agrees on. checked 5 August 2026.
Before publishing a disagreement this size, the obvious suspicion is that our own method is wrong — that the terrain model is reading a rock face beside the road rather than the road. Two checks say otherwise. Everywhere else the same method lands close: across 35 comparisons on 32 other-sourced summits — 32 against an altitude somebody else published and 3 against this site’s separate measurement from the Mallorca 312 GPX tracks — the median disagreement is 2 m and 33 of them fall inside 10 m. And sampling a 400-metre square around the Coll dels Reis node finds a smooth saddle between 694 m and 746 m — no notch, no cutting, nothing for a terrain model to misread.
The same method, checked against everyone else's altitudes
| Summit | Ours | Theirs | Difference | Compared against |
|---|---|---|---|---|
| coll de sa Bastida | 298 m | 298 m | 0 m | OpenStreetMap node 1621679173 |
| Coll de sa Creu | 322 m | 322 m | 0 m | OpenStreetMap node 1360452883 |
| Coll de Sóller | 497 m | 497 m | 0 m | OpenStreetMap node 26730946 |
| coll de Tofla | 261 m | 261 m | 0 m | OpenStreetMap node 9859055963 |
| Coll dels Tords | 276 m | 276 m | 0 m | OpenStreetMap node 306616949 |
| Coll des Pescadors | 374 m | 374 m | 0 m | OpenStreetMap node 9192893096 |
| es Grau | 331 m | 331 m | 0 m | OpenStreetMap node 9694410630 |
| pas de na Bou | 36 m | 36 m | 0 m | OpenStreetMap node 6357014881 |
| Coll d'en Capó | 360 m | 360 m | −0 m | OpenStreetMap node 3397946917 |
| Can Bleda | 223 m | 224 m | −1 m | OpenStreetMap node 12709284239 |
| Coll d'en Claret | 499 m | 498 m | +1 m | OpenStreetMap node 2518205347 |
| Coll d'Orient | 499 m | 498 m | +1 m | OpenStreetMap node 1280164782 |
| coll de la Creueta | 208 m | 207 m | +1 m | OpenStreetMap node 384926035 |
| Coll de n'Esteve | 239 m | 238 m | +1 m | OpenStreetMap node 8077008023 |
| coll de s'Heura | 323 m | 322 m | +1 m | OpenStreetMap node 313984662 |
| pas de can Bregat (Pollença) | 23 m | 24 m | −1 m | OpenStreetMap node 6357014877 |
| Coll de Morell | 196 m | 194 m | +2 m | OpenStreetMap node 3068188469 |
| Can Costa | 426 m | 424 m | +2 m | OpenStreetMap node 11810288320 |
| Coll de sa Batalla | 578 m | 576 m | +2 m | OpenStreetMap node 1280164795 |
| Coll de sa Coma | 224 m | 226 m | −2 m | OpenStreetMap node 12709284240 |
| coll de sa Teulera | 156 m | 158 m | −2 m | OpenStreetMap node 306618681 |
| Coll des Grau | 466 m | 468 m | −2 m | OpenStreetMap node 3978212050 |
| Galilea | 408 m | 410 m | −2 m | OpenStreetMap node 26492767 |
| Túnel de Monnàber | 873 m | 875 m | −2 m | OpenStreetMap node 9700247247 |
| Coll des Racó | 226 m | 229 m | −3 m | OpenStreetMap node 271800720 |
| Coll de sa Gramola | 357 m | 360 m | −3 m | OpenStreetMap node 1690751012 |
| Coll des Pi | 237 m | 240 m | −3 m | OpenStreetMap node 1626869215 |
| pas de l'Escala de l'amo en Joan Cerdà | 134 m | 130 m | +4 m | OpenStreetMap node 6357014817 |
| Coll d'Honor | 555 m | 550 m | +5 m | OpenStreetMap node 356409154 |
| Coll de Femenia | 537 m | 532 m | +5 m | our own measurement from the Mallorca 312 official GPX |
| Coll de sa Creu | 378 m | 384 m | −6 m | OpenStreetMap node 1360452952 |
| Puig Major (MA-10 from Gorg Blau) | 873 m | 866 m | +7 m | our own measurement from the Mallorca 312 official GPX |
| Coll de sa Gramola | 357 m | 348 m | +9 m | our own measurement from the Mallorca 312 official GPX |
| Coll de Femenia | 537 m | 515 m | +22 m | OpenStreetMap node 765875444 |
| Coll dels Reis | 723 m | 682 m | +41 m | OpenStreetMap node 1123226364 |
How we got this Each summit’s altitude from the IGN 5 m terrain model, minus the independently published figure for the same point. Passes only — a pass node is on the road, a peak node is not. Rows sorted by the size of the disagreement.Computed from OpenStreetMap elevation tags on the pass nodes, and this site’s separate measurement of three summits from the Mallorca 312 organiser’s official GPX tracks.Computed 5 August 2026.
The German sources also settle a length this page never had an answer for. German Wikipedia gives the road «12,5 Kilometern Länge» against our 10.0 km, and the two were never in conflict: one measures the road and the other the climb. That is checkable here rather than assertable, because our record holds both sides of this pass, and the Ma-2141 runs through both.
How we got this The two measured sides of this pass between them cover the whole Ma-2141. Adding the metres each carries of that one road gives the length of the road itself, which is a different quantity from the length of the climb and is the one other publishers state.Computed from Our measured road breakdown for both sides of the Coll dels Reis: 10019 m of Ma-2141 climbing from Sa Calobra, 2688 m of it on the Lluc side.Computed 7 August 2026.
And the same article publishes 682 m as a height difference over those 12.5 km, where every other source publishes it as an altitude. On this particular road one number can pass for both, because the road starts 15 m above the sea — which is exactly what our two figures, 723 m and 708 m, keep apart.
So the measurement stands, and the explanation turns out to be dull, which is the best kind. The published figures describe a climb that stops about 600 m short of the top of the road.
How we got this At 9.4 km from the port — the length almost everyone publishes — our series reads 670 m, twelve metres from the 682 m most sources give as the summit. The road then climbs for another 600 m to its actual high point at 723 m. Two descriptions of the same road that stop in different places, not two contradictory measurements of the same thing. It also explains why the sources that measure the road to its end — quaeldich’s 10.1 km and 728 m — land on our figures rather than on the shorter set.Computed from Our 10 m elevation series for the Ma-2141, against the 9.4 km / 682 m figures in common circulation.Computed 7 August 2026.
The road
The Ma-2141 was designed by Antonio Parietti, and the two sources we checked put its construction a year apart — opened 1933 and built 1932.
Two of the three say 1932 and one says 1933, and none shows its own source. Build and opening are different events, which is the likeliest explanation, and the German edition’s «geplant und 1932 ausgeführt» is the wording that says so most directly.The nus de sa corbata, as described by the sources that describe it
| Angle of the loop | Published by | In their words |
|---|---|---|
| 270° | Wikipedia (en) | A “270° spiral bridge called the tie knot”. |
| 270° | PJAMM Cycling | Describes the “Nus de Sa Corbata” as a 270-degree turn at about kilometre 9.5 — which is also where our own path check puts the crossing, 9.46 km up. |
| 360° | Dangerous Roads | Calls the nus de sa corbata “a curve of 360 degrees” — and, in the same page, describes the road “turning 270 degrees and going under itself”. The two figures are the source’s own, not a disagreement between sources. |
| 270° | Wikipedia (de) | Calls it a «270°-Kehrschleife», and is the only source here that says why it is there: at Sa Moleta, Parietti «sah nur die Möglichkeit» of losing the height that way. |
| 270° and 360° | quaeldich.de | Its ascent text ends the climb «500 m vor der berühmten 270-Grad-Kurve am Coll dels Reis»; its own photo caption on the same page calls the same corner «die berühmte 360-Grad-Serpentine». The second source in this table to publish both figures itself. |
| 255° | Our measurement (bend detection on the OpenStreetMap centreline) | The largest single turn our counter finds, with its apex 64 m from the bridge the map tags on layer 2. It measures the swing of the road’s centreline, which is not quite the same question as how far the structure carries you round. |
Four publishers say 270° and none of them says how it was arrived at. The two that reach 360° both print 270° on the same page as well, which makes this one figure with a loose second one attached rather than a genuine split. Our own reading of the centreline is 255°, and the difference between that and 270° is what the road does either side of the bridge rather than a contradiction. checked 5 August 2026.
The reason the loop exists at all is arithmetic rather than showmanship: the road had to lose
height in a place with no room to lose it, and taking the alignment under itself bought the
distance without boring through anything. OpenStreetMap corroborates that much directly — the
loop is mapped as a bridge on layer=2, the road passing over its own alignment.
Dangerous Roads adds that the road was cut by hand, moving 31,000 m³ of rock, and was deliberately designed so that no tunnel was needed anywhere on it. The second half of that does not survive contact with the map: OpenStreetMap records a short tunnel on the Ma-2141. It is 28 m long, which is why the profile above says the elevation is interpolated across it — and it is worth holding next to the modern Ma-10, where the Puig Major climb from Sóller runs 409 m of its length underground.
How we got this The measurement flags every way tagged as a tunnel or a bridge and interpolates elevation between the portals. On this climb that flagged one tunnel — the metres are reported on the profile — and several bridges, one of them carrying the road over its own alignment on layer 2, which is the tie knot.Computed from OpenStreetMap way 667742888, tagged tunnel=yes on ref Ma-2141, and our own measured series.Computed 7 August 2026.
Everybody says 26 hairpins. It depends what you count.
Twenty-six is the number this road is famous for. Cycling UK prints it of the descent — “The fast, twisting 10km descent includes 26 hairpins” — and it is the figure most guides to the Ma-2141 repeat. It is also not reproducible, because nobody publishes what they counted. So we counted the bends ourselves, from the same OpenStreetMap centreline the gradients above are sampled along, under a rule written down before it was run: a turn of at least 135° taken inside 150 m of road.
The bends of the Ma-2141: our three counts and the 3 published ones
| Counting | How many |
|---|---|
| Turns of at least 135° taken inside 150 m of road — our published rule | 15 |
| Turns of at least 135°, over any length of road | 23 |
| Turns of at least 90°, over any length of road | 28 |
| 26 hairpins, plus the loop where the road passes under itself — Cycling UK | 26 |
| 14 hairpins, in a stated stretch of road — PJAMM Cycling | 14 |
| 12 hairpins, over the whole road rather than the climb — Wikipedia (de) | 12 |
Our strict rule gives 15 for the whole climb, and 23 turns reach 135° over any length; 28 bends reach 90°. The two published counts that describe the whole road bracket our strict figure from both sides — 12 below it, 26 above — and they differ from each other by more than a factor of two on one road, which is what a count without a stated rule looks like when you try to reproduce it. PJAMM’s 14 is a different kind of claim: it names the stretch it counted, and it comes back exactly. checked 6 August 2026.
The two published counts that describe the whole road bracket our strict count from both sides, and they are further from each other than either is from us: German Wikipedia counts 12 over a stretch it gives as 14 km, and Cycling UK counts 26 over one it gives as 10 km. One road, two figures, a factor of two. That is not a rebuttal of either — it is the reason three guides to the Coll de Sóller publish three different numbers for one pass. A count without a stated rule cannot be checked, only believed, and this is a road people count while descending it.
How we got this Bearings every 5 m along the centreline; consecutive turns in the same direction accumulated into one bend; a bend ends after 25° of counter-turn.Computed from OpenStreetMap centreline (ODbL), and the source Cycling UK publishes.Computed 6 August 2026.
There is one place on this island where a completely independent claim says “there is a corner
here”, and it is on this road. OpenStreetMap records the nus de sa corbata as a bridge on
layer=2 — the road over its own alignment — which is topology, with no angle in any tag. The
counter, which has never seen that tag, puts its single biggest turn of the whole climb in the
same hundred metres: The Ma-2141 crosses over itself 9.46 km up the climb. The counter's nearest turn is 255° and its apex is 64 m away, counted as a hairpin.
How we got this Every edge of the derived path is checked for a bridge tag on layer 2 or above — the road over its own alignment. The apex of the nearest detected turn is compared with the crossing point. Nothing in the OpenStreetMap tags states an angle, so the two claims are independent.Computed from OpenStreetMap way 25073139, tagged bridge on layer 2, against our own bend detection on the same centreline.Computed 6 August 2026.
That check, the rule behind it and every other road on the island measured against it are on the twistiest roads on Mallorca — which also finds that this climb is not the twistiest here, on any of the three counts. It ranks 4 of 116.
Whether you can ride it, and what is changing
Source Consell de Mallorca · checked 6 August 2026
That was the emergency. The programme that follows it is the one you would feel through the bars. The Consell de Mallorca budgeted €3 million to renew the road surface completely and to fit new barriers and updated signage over the 13.3 kilometres from the Coll dels Reis down to the port of Sa Calobra, with work announced to start in September 2025.
Source Ultima Hora · checked 7 August 2026Reported 2 January 2025. The 13.3 km is the whole road, not the climb we measure — our ascent is the 9.4 km from the port up. The article states no completion date and says nothing about the 10-tonne limit.
In September 2025 the Consell said the reform of two Serra de Tramuntana roads on its roadmap — the Coll de Sóller and Sa Calobra — would carry on into the following year. So a rider arriving in 2026 should expect the Ma-2141 to be a road with work on it rather than a road that has been finished — and should expect the same of the Coll de Sóller, which is the other name in that sentence.
Source Ultima Hora · checked 7 August 2026Reported 3 September 2025: «Para el próximo año está previsto seguir adelante con la reforma de dos carreteras de la Serra de Tramuntana que están dentro del libro de ruta: las del Coll de Sóller y Sa Calobra.» “El próximo año” from a September 2025 report is 2026, which is now. It is a statement of intent, not of progress: neither we nor it says how much of either road is done.
What is new is that this specific road is being wired. 210 cameras — 105 with number-plate readers and vehicle counting, 105 for surveillance — and 32 control zones in the Serra de Tramuntana, plus 20 sound meters distributed across the whole island, for €1 million. The first 64 devices are going in on the Lluc–Sa Calobra road and at the Caimari entrance, and the system is to be fully operational by October 2026.
Source Ultima Hora · checked 6 August 2026Reported 25 July 2026. The stated purposes are illegal racing, noise, and counting vehicles entering from outside Mallorca ahead of a vehicle-entry law still in parliament; the article notes the AI can distinguish vehicle types and detect bicycles.
For a cyclist the relevant part is not enforcement but arithmetic: the stated purpose includes counting the vehicles that enter, ahead of a vehicle-entry law still in parliament. The Lluc–Sa Calobra road is where the counting starts.
The air at the top, and what the way down feels like
This descent ends at the water, and the only way home is back up through 708 m of it. In January the top of that road runs 5.9 °C colder than the sea you turned round at — a gap you climb into rather than descend through.
You finish this climb 708 m above where you started it, and the air knows. In January — the coldest month at this top — the foot of the climb averages 14.3 °C around 12:00 while the top averages 8.4 °C. That is a different set of clothes at each end: 5.9 °C bought with 708 m of climbing. In July the same top sits at 26.1 °C.
Then you turn round. Going down at 30–50 km/h you make your own wind, and at 08:00 in January the air at this top — 5.4 °C standing still — feels like 0.6 °C at the slower speed and -0.8 °C at the faster one. Ride it at 12:00 instead and the same descent reads 3.3 °C to 4.5 °C — the index speaks about the middle of the day here in two months of the twelve. Whatever you would wear on the way up, the way down is the reason the gilet is in the pocket.
The air at Coll dels Reis, month by month
| Month | Foot, 08:00 | Top, 08:00 | Top, 12:00 | Descending, 08:00 |
|---|---|---|---|---|
| January | 7.8 °C | 5.4 °C | 8.4 °C | -0.8 °C – 0.6 °C |
| February | 7.2 °C | 6.2 °C | 9.9 °C | 0.3 °C – 1.6 °C |
| March | 10.0 °C | 7.2 °C | 10.8 °C | 1.7 °C – 2.9 °C |
| April | 11.9 °C | 9.1 °C | 12.9 °C | 4.3 °C – 5.4 °C |
| May | 15.9 °C | 13.1 °C | 16.7 °C | — |
| June | 20.7 °C | 18.4 °C | 21.9 °C | — |
| July | 23.3 °C | 21.6 °C | 26.1 °C | — |
| August | 22.8 °C | 21.3 °C | 25.6 °C | — |
| September | 20.1 °C | 17.7 °C | 21.7 °C | — |
| October | 15.0 °C | 12.9 °C | 16.6 °C | — |
| November | 12.1 °C | 9.3 °C | 12.0 °C | 4.5 °C – 5.7 °C |
| December | 9.3 °C | 7.5 °C | 10.8 °C | 2.1 °C – 3.3 °C |
How we got this Hourly readings from two stations 1,007 m apart, paired so that every figure is a difference between two instruments reading the same instant. The temperature at 723 m is read between them and never projected past the upper one. The descent column is the wind chill index at 30–50 km/h of the rider's own air, printed only inside the range its publisher states it holds, and taken at 08:00 because that is the hour it speaks about at every top on this island.Computed from 27,536 paired hours at 8 m and 1,015 m, 2018–2021.Computed 17 August 2026.
Two thermometers 1,007 m apart define a line, not a surface, so a top on the north coast reads a figure taken across the island. Every figure here is measured under an open sky: where the mountain's own shadow lies, read the table as the warm version of the morning. What the temperatures do across the whole island, and what to put in the pocket, is on the kit guide.
Getting there, and getting out
The Ma-2141 leaves the Ma-10 near Escorca and dead-ends at the cove. There is no loop, no alternative descent, and no road along the coast in either direction. Every rider who goes down either comes back up or leaves by sea, and every car and coach that goes down does the same thing on the same two lanes — which is the entire reason this road has been in the island’s traffic politics for a decade.
The other side of the pass is a different ride entirely: from Lluc it is 13.8 km at 1.8% for 248 m, because it runs almost the whole way on the Ma-10 and only switches onto the Ma-2141 at the end. Against the island median of 192 m, the haul out of the cove is a big climb — 2 of 75 — and it is the reason anybody is here at all.
If you are basing a trip around the Tramuntana rather than staying in it, a hire car is what makes the Serra reachable from Palma; if you would rather wake up at the foot of these roads, Pollença, Sóller and Alcúdia are the three bases that put Sa Calobra inside a morning.
Verified Every figure on this page was re-checked against its source on 5 August 2026. Next review due within 120 days.
Sa Calobra on one page
The per-kilometre splits and the gradient profile, formatted to fold into a jersey pocket. Sent once, plus a note when the measurement is revised.