Who Were the Huns? DNA Traces Attila's Elite to the Xiongnu Nobility, but Not His Empire

Hero image: AI generated illustration of a mixed group of riders on the Pannonian plain in the 5th century. It is not a reconstruction of real individuals. The standing stones in the background are artistic licence and do not belong to the Hungarian plain.

In the 370s a people the Romans had never heard of crossed the Volga and smashed the Alans and the Goths north of the Black Sea. Within two generations the Huns ruled from the steppe to the Rhine, extracted thousands of pounds of gold from Constantinople, and, under Attila, invaded Gaul and Italy. Then, almost as fast, they vanished. Attila died in 453, his sons lost the battle of Nedao the next year, and the Hun empire dissolved into its subject peoples. The Huns left no texts of their own, no cities and only a few dozen words of their language. Ever since, one question has hung over them: where did they come from?

The most famous answer is more than 250 years old. In the 1750s the French orientalist Joseph de Guignes proposed that the Huns were the Xiongnu, the nomadic empire that fought Han China across the Mongolian steppe from the 3rd century BC and gave the Great Wall much of its purpose. The names look alike, and a Sogdian merchant's letter written around 313 AD calls the Xiongnu who sacked the Chinese capital Luoyang by the name Xwn. Sceptics answered that more than two centuries and 6,000 kilometres separate the fall of the northern Xiongnu from the Huns on the Volga, and that a name proves nothing about ancestry.

Ancient DNA can test this. And the answer turns out to be more interesting than either camp expected.

The first genomes: a steppe mixture

The first large genomic survey of the Eurasian steppe, by de Barros Damgaard and colleagues in 2018, sequenced 137 ancient individuals, including people buried in Hun-period graves in the Tian Shan mountains of Kyrgyzstan and Kazakhstan. They found that the Xiongnu were themselves a mixed population, formed when East Asian groups of the Mongolian plateau merged with western steppe nomads related to the Scythians and Saka. As this Xiongnu-related ancestry spread west, it mixed again with local steppe populations. The Central Asian Huns of the 2nd to 5th centuries looked like the result: Saka and Sarmatian-like people with a variable East Asian share.

Hungarian teams then went straight to the Carpathian Basin, the heart of Attila's empire. Neparáczki and colleagues typed the Y chromosomes of Hun, Avar and Conquest-period nomads in 2019, and Maróti and colleagues published whole genomes from all three periods in 2022. Their handful of Hun-period samples included individuals who were genetically almost entirely East Asian, closest to the late Xiongnu, together with others who carried Sarmatian, Germanic or local ancestry. Among the Hun-period men, the Y lineages included R1a-Z93 and Q, both common on the eastern steppe. The sample was small, but the message was already there: the Huns were not one people.

The 2025 study: Xiongnu nobility, but no Xiongnu nation

In February 2025 a team led by Guido Alberto Gnecchi-Ruscone, Zsófia Rácz and Zuzana Hofmanová, working within the HistoGenes project of the Max Planck Institute for Evolutionary Anthropology and Eötvös Loránd University, published the largest test so far in PNAS. They analysed 370 individuals spanning eight centuries: Xiongnu-period burials from the Mongolian steppe, 3rd and 4th century graves from Kazakhstan, and 5th and 6th century Hun and post-Hun period burials from the Carpathian Basin, 35 of them newly sequenced.

The headline is a double result. First, the authors write that they "find no evidence for the presence of a large eastern/steppe descent community among the Hun- and post-Hun-period Carpathian Basin population." Only a few Hun-period individuals carried substantial East Asian ancestry, and the population of the Hun realm in Europe was, in their words, genetically highly heterogeneous.

Second, and this is what made the headlines, a few of those East Asian individuals share long identical-by-descent (IBD) segments of DNA with members of the late Xiongnu imperial elite in Mongolia. Shared segments that long are only inherited from relatively recent common ancestors. One of the links goes to an individual buried in the largest terrace tomb ever found in a Xiongnu context. Another of the connected individuals is a woman aged 35 to 50, buried in the first half of the 5th century at Pusztataskony in Hungary with gold earrings and an artificially elongated skull. As Gnecchi-Ruscone put it: "It came as a surprise to discover that a few of these Hun-period individuals in Europe share IBD links with some of the highest-ranking imperial elite individuals from the late Xiongnu Empire."

So de Guignes was half right. There is a genealogical thread running from the Xiongnu aristocracy of Mongolia to Attila's Hungary. But it runs through a few families, not through a people. Hofmanová: "Although the Huns dramatically reshaped the political landscape, their actual genetic footprint, outside of certain elite burials, remains limited."

Global25 cannot detect IBD segments, so we cannot repeat that part. What we can do is measure, individual by individual, how much East Asian ancestry the people of the Hun period actually carried, and compare them with the Xiongnu, the Central Asian Huns, the Avars who came two centuries later, and the Hungarians.

Our data

We pulled every individual dated between about 370 and 480 AD from Hun-period contexts in the Carpathian Basin out of the dated Global25 ancient file: 49 people, including the Hun-period samples of Maróti 2022, the Late Sarmatian and early Hun period cemeteries of the Danube, Tisza and Transtisza regions, and the three samples labelled as East Asian outliers. Around them we placed eight comparison groups: late Xiongnu from Mongolia (11), Xianbei-Hun period people from Berel in the Kazakh Altai (12), Tian Shan Huns from Kyrgyzstan (22), the Hun elite burial from Kurayly in Kazakhstan, Carpathian Sarmatians before the Huns (118), people buried in the Middle Tisza region after Attila (13), Early Avars (106) and conquering Hungarians of the 10th century (90).

Each genome was modelled as a mixture of four deep sources: local Carpathian Iron Age people (Hungarian La Tene), Germanic people of the Wielbark culture in Poland (the milieu of the early Goths), Sarmatian steppe nomads, and Neolithic hunter-gatherers of northern Mongolia as a clean Northeast Asian pole. The model is well conditioned (condition number 13.2; the Northeast Asian pole sits 0.52 to 0.61 from the three others). The two European sources are close to each other (0.076), so the split between local and Germanic is soft. The Northeast Asian share, the number this article is about, is robust: dropping the Germanic source changes it by less than half a percentage point on average.

One subtlety matters. Sarmatians themselves carried some East Asian ancestry (roughly 15 percent in a model without a Sarmatian source). The Northeast Asian figure below is therefore the East Asian ancestry beyond what an ordinary Sarmatian would bring. It measures the eastern newcomers, not the background of the steppe.

The 49 people of the Hun period

The distribution is stark. Of the 49 Hun-period individuals:

• 36 (73 percent) carry less than 5 percent Northeast Asian ancestry, and most carry none that the model can detect. They are mixtures of local Carpathian, Germanic and Sarmatian ancestry, indistinguishable from the people who lived in the Basin before the Huns arrived.
• 8 carry between 5 and 20 percent, mostly in Late Sarmatian to early Hun period cemeteries of the Upper Tisza. These look like local families with an eastern great-grandparent or two.
• 5 carry more than 20 percent. One individual (KMT-2785) is a genuine mixture, about one third Northeast Asian. The other four are something else entirely.

The four strongly eastern individuals (ASZK-1, PTL013, VZ-12673 and HUN001) carry between 73 and 84 percent Northeast Asian ancestry, as much as the late Xiongnu themselves (median 82 percent in our 11 Mongolian samples). HUN001 has low coverage (12 percent of sites), but its signal is so far from the European cluster that noise cannot explain it.

Who were the four?

We compared these four genomes with every candidate source in the dataset, deliberately leaving out the Moriopoulos "Hun elite" groupings, which contain the same individuals and would make the test circular.

The result lines up with the IBD findings from a completely different angle. The four sit between 0.03 and 0.07 from the late Xiongnu average and from the eastern elite outliers of Central Asia, but 0.19 to 0.31 from the ordinary Hun-period populations of the Tian Shan and the Altai. In other words, they do not resemble the typical mixed steppe population that the Huns supposedly picked up on their way west. They resemble the Xiongnu core.

A proximal model makes the same point. Offered late Xiongnu, Carpathian Sarmatians and local La Tene people as sources (condition number 13.3), all four individuals come out as 100 percent late Xiongnu, with no local admixture needed. So does the Kurayly elite burial in Kazakhstan. The Tian Shan Huns, by contrast, model as 44 percent late Xiongnu and 56 percent Sarmatian-like.

Two readings are possible, and the data cannot separate them. Either these four were themselves migrants, or the children of migrants, who arrived from the east within a generation or two. Or they belonged to families that married almost only among themselves for generations while crossing the steppe, the way ruling dynasties often do. Either way, the eastern element in Attila's Europe was a thin, high-status layer that had not mixed with its subjects.

Is a zero really a zero?

Our claim that 36 of the 49 carry no eastern ancestry rests on the model being able to see small amounts. So we tested it. We took the average of those 36 people and spiked in known amounts of Northeast Asian ancestry, first as the pure Mongolian pole, then as real late Xiongnu genomes, and re-ran the model.

A 2 percent spike is recovered as 1.4 percent, and a spike of real Xiongnu genomes behaves the same way. The same test on single high-coverage individuals (SEI-6, 77 percent coverage) gives the same floor. To check for the opposite error, false positives, we ran the model on 161 Roman, Late Antique, Gothic-period and Langobard individuals from Croatia, Serbia, Hungary and Italy, people with no reason to carry East Asian ancestry. 99 percent of them score below 3.2 percent. The honest detection floor is about 2 to 3 percent: anything above 5 percent in Figure 2 is real, and the zeros mean "less than about 2 percent".

After Attila

One more group deserves attention. In cemeteries of the Middle Tisza region dated to around 485, a generation after Attila's death and the collapse of his empire, 5 of the 13 sequenced people carry more than 50 percent Northeast Asian ancestry. That is a much higher share than in the Hun period itself. The eastern families did not all leave with Attila's sons. Some apparently stayed on the Great Hungarian Plain under Gepid rule and kept a distinct ancestry for at least another generation, a pattern the 2025 study also describes for post-Hun period communities.

Huns versus Avars: two very different migrations

The contrast with the Avars is the clearest result in Figure 1. The Avars arrived in the Carpathian Basin in 568, about a century after the Hun empire fell. Of our 106 Early Avar individuals, 75 carry more than 20 percent Northeast Asian ancestry, with a median of 36 percent, against a median of practically zero for the Hun period. Walter Pohl, a historian involved in the HistoGenes project, summarised the difference: the Avars "came directly to Europe after their East-Asian empire had been destroyed by the Turks, and many of their descendants still carried considerable East Asian ancestry until the end of their rule in circa 800."

The Huns did not. Whatever group set out from the eastern steppe in the 2nd or 3rd century took roughly two hundred years to reach the Danube. By then the empire they ran was a confederation of Sarmatians, Alans, Goths, Gepids, Romans and many others, and its genetic profile was that of its subjects. The 5th century writer Priscus, who visited Attila's court, described Greeks, Goths and Latin speakers living among the Huns. The genomes agree with him.

And the Hungarians?

Hungarian medieval chronicles, above all Simon of Kéza's Deeds of the Huns and Hungarians in the 1280s, presented the Magyars as Attila's heirs, and the idea remains popular. Two separate questions hide in it.

Did the Hungarian conquerors of the 10th century carry eastern ancestry? Yes. In our 90 Conquest period individuals, the median is 7 percent Northeast Asian and 33 of them carry more than 20 percent. Maróti and colleagues modelled the Conqueror core as an Ugric-related population from the Ural region that had mixed with Sarmatians and with descendants of the late Xiongnu, so there is a distant Xiongnu thread there too. But it arrived with a different migration, four centuries after Attila.

Do modern Hungarians carry measurable Hun ancestry? In a three-source model (local Carpathian, Germanic, Northeast Asian), modern Hungarians come out at 1.7 percent Northeast Asian. Their neighbours give 2.3 percent (Romanians), 1.2 percent (Croatians) and 1.0 percent (Austrians). All of these sit inside the noise band we measured. Genetically, modern Hungarians are Central Europeans, and any specifically Hunnic ancestry is too diluted to identify with Global25.

So who were the Huns?

Not a nation, and not a race. The best reading of the DNA is a political label worn by a very mixed population, led by a core of families whose ancestors had belonged to the highest ranks of the Xiongnu empire in Mongolia. That core travelled for two centuries and thousands of kilometres, and some of its members still looked genetically like Xiongnu nobles when they were buried on the Hungarian plain. Around them, the people who rode under Attila's banner were Sarmatians, Goths, Gepids, provincial Romans and the local population of the Carpathian Basin, who shared the Huns' customs, including skull binding and eastern-style horse gear, without sharing their ancestry. When the core families lost power, the "Huns" simply dissolved back into the peoples they had been made of.

Global25 coordinates used in this article

Group averages computed from the dated individual ancients file (Global25 by Davidski). Paste directly into the Vahaduo tool.

Carpathian_Hun_period_low_EastAsian_(n=36),0.125142,0.126546,0.037586,0.018106,0.023688,0.006763,0.002996,0.003269,-0.002352,0.004434,-0.001001,0.003530,-0.007342,-0.002749,-0.000422,-0.000435,0.002144,0.000912,0.004197,-0.000889,-0.003601,0.000921,0.001191,0.002460,-0.002325
Carpathian_Hun_period_EastAsian_rich_(n=4),0.047237,-0.332586,0.057134,-0.011224,-0.049471,-0.019801,0.006815,0.014653,0.006033,0.001549,-0.030326,-0.000037,0.005501,-0.005367,0.000068,0.003944,0.003586,-0.006714,0.004745,0.010224,-0.015941,-0.000185,-0.008165,0.003193,-0.000629
Carpathian_Hun_period_KMT-2785_(n=1),0.094473,-0.074134,0.059585,0.015827,-0.001231,-0.001673,0.007990,0.001615,0.001841,-0.004556,-0.015589,-0.000899,-0.001487,-0.004129,0.002036,0.014320,0.009127,0.004687,-0.000503,0.007879,-0.004617,-0.001731,-0.002835,0.002892,0.000838
Mongolia_Late_Xiongnu_(n=11),0.055877,-0.304289,0.065825,-0.006401,-0.059843,-0.023503,0.009507,0.014915,-0.001153,0.002833,-0.026838,0.000068,0.001324,-0.009333,0.003541,0.004255,-0.001553,-0.003017,0.002788,0.009675,-0.015847,-0.001900,-0.014969,-0.001260,0.001208
Kyrgyzstan_TianShan_Hun_(n=22),0.087075,-0.071272,0.037044,0.045381,-0.042050,0.007023,0.002564,0.003745,-0.017478,-0.024850,-0.009035,-0.001417,0.002027,-0.021913,0.015417,0.008341,-0.003858,0.001820,-0.000388,0.003456,-0.014208,0.001113,-0.002090,0.002843,-0.000278
Carpathian_Sarmatian_pre-Hun_(n=118),0.120643,0.110185,0.042151,0.028826,0.020510,0.008629,0.003298,0.003242,-0.002602,-0.002158,-0.001094,0.002162,-0.004341,-0.002616,0.000766,0.001773,0.002293,0.000412,0.003167,-0.000288,-0.003178,0.002088,0.000169,0.003794,-0.001275
Carpathian_Early_Avar_(n=106),0.086377,-0.092289,0.048214,-0.005034,-0.022457,-0.015302,0.007609,0.011024,-0.002576,0.004052,-0.011111,0.001026,-0.003304,-0.000422,-0.000338,-0.000278,0.000857,-0.000893,0.003424,0.004149,-0.010531,-0.000264,-0.005704,0.002103,0.000445

Methods in brief

Global25 scaled coordinates were merged with first occurrence priority from Davidski's population averages, modern averages, the Moriopoulos 2026 collection (all averages, no sims) and the dated individual ancients file; scale compatibility was checked on shared populations (distance 0.009 between the published Mongolia_North_N average and the mean of its individuals). Hun-period individuals were selected by context label and date (370 to 481 AD). Distances are Euclidean over 25 dimensions. NNLS models use a sum to one constraint enforced by a heavily weighted extra equation, with results normalised. The deep model uses Hungary_IA_LaTene, Poland_Weklice_WielbarkCulture_Roman, Russia_Sarmatian and Mongolia_North_N (condition number 13.2). The proximal model uses the mean of 11 late Xiongnu individuals, the mean of 118 Carpathian Sarmatians and Hungary_IA_LaTene (condition number 13.3; the two European sources are 0.055 apart, so their split is uncertain). Spike-in tests add a known fraction of a source to the target and re-fit. The false positive test uses 161 Roman, Late Antique, Gothic-period and Langobard individuals. Duplicate libraries were removed. G25 results are a proxy; the published qpAdm and IBD results take precedence where they exist.

References

1. Genomics Gnecchi-Ruscone GA, Rácz Z, Liccardo S, et al., Hofmanová Z (2025). Ancient genomes reveal trans-Eurasian connections between the European Huns and the Xiongnu Empire. Proceedings of the National Academy of Sciences 122(9). doi.org/10.1073/pnas.2418485122
2. Genomics Maróti Z, Neparáczki E, Schütz O, et al. (2022). The genetic origin of Huns, Avars, and conquering Hungarians. Current Biology 32(13): 2858 to 2870. doi.org/10.1016/j.cub.2022.04.093
3. Y-DNA Neparáczki E, Maróti Z, Kalmár T, Maár K, Nagy I, Raskó I, et al. (2019). Y-chromosome haplogroups from Hun, Avar and conquering Hungarian period nomadic people of the Carpathian Basin. Scientific Reports 9. doi.org/10.1038/s41598-019-53105-5
4. Genomics de Barros Damgaard P, Marchi N, Rasmussen S, et al. (2018). 137 ancient human genomes from across the Eurasian steppes. Nature 557: 369 to 374. doi.org/10.1038/s41586-018-0094-2
5. Press release Max Planck Society (2025). Origin and diversity of Hun Empire populations. mpg.de
6. Data Global25 PCA coordinates (Davidski) and the Moriopoulos 2026 collection of Global25 averages.