Oxford Nanopore vs PacBio HiFi: Choosing a Long-Read Sequencing Platform

Every genome project eventually forces the same expensive decision: Oxford Nanopore or PacBio HiFi? The choice locks in your basecaller, your read-length distribution, your per-base error profile, and often your entire assembly strategy weeks before a single contig exists. Pick wrong and you spend the project fighting the data instead of the biology. This post breaks down the real tradeoffs between ONT's R10.4.1 chemistry with Dorado and PacBio's HiFi reads from the Revio instrument, so you walk into your project knowing which platform fits the question you are actually asking.

Two Different Ways to Read a Long Molecule

Both platforms sequence single native DNA molecules that are tens of kilobases long, but they get there through completely different physics, and that difference drives everything downstream.

PacBio uses single-molecule real-time (SMRT) sequencing. A circularized library (a SMRTbell) is read around and around by a polymerase, so the same insert is sequenced many times. Each pass is noisy, but averaging many passes into a circular consensus sequence (CCS) collapses the random error down to something extraordinarily clean. That consensus read is what everyone calls a HiFi read.

Oxford Nanopore threads a single DNA strand through a protein pore and measures the ionic current disruption as bases pass through. The R10.4.1 pore has a longer sensing region that reads two contact points at once, which is exactly why modern ONT accuracy jumped so dramatically over the older R9.4.1 chemistry. There is no consensus-by-repetition step here — the accuracy comes from better chemistry and, critically, from the neural-network basecaller Dorado interpreting the raw squiggle.

Accuracy: The Number Everyone Argues About

Accuracy is where the marketing gets loudest, so let's be precise about what the numbers mean. Base quality is reported on the Phred scale, where Q20 is 99% accurate, Q30 is 99.9%, and Q40 is 99.99%.

PacBio HiFi reads are defined by a quality floor: a read only counts as HiFi once its consensus clears roughly Q20, and typical modal accuracy sits around Q30 and above. That consistency is the whole selling point — you can trust nearly every base without polishing.

ONT R10.4.1 with Dorado in its super-accurate (sup) mode reaches modal single-read accuracy of Q20+ for standard simplex reads, a genuine leap from where Nanopore sat a few years ago. ONT also offers duplex reads, which pair the complementary strand's signal from the same pore pass into a second, higher-confidence consensus — duplex reads routinely clear Q30, closing much of the gap with HiFi, but duplex yield is a fraction of total throughput, so most production runs are still reported at simplex accuracy. The important nuance is the shape of the error distribution, not just the average.

Read Length: Where Nanopore Pulls Ahead

This is the axis where the platforms genuinely diverge, and it is often the deciding factor.

PacBio HiFi reads are capped by the consensus process. To read a molecule many times, the insert has to be short enough for multiple polymerase passes — in practice that means HiFi inserts land around 15 to 25 kb. Long, accurate, but bounded.

ONT has no such ceiling. A standard ligation library routinely yields read N50 values well above 30 kb, and with careful high-molecular-weight extraction you can push into ultra-long territory where reads exceed 100 kb and the tail stretches past a megabase. Those ultra-long reads are the ones that span segmental duplications, tandem repeat arrays, and centromeric satellite DNA that nothing else can bridge. The telomere-to-telomere human assemblies leaned hard on ONT ultra-long reads for exactly this reason.

If your genome is repeat-heavy, or you specifically need to resolve structural variants and complex loci, read length is not a nice-to-have — it is the entire game, and ONT wins it.

Throughput, Scale, and Cost

Cost per gigabase changes fast, so treat exact figures as moving targets, but the structural differences are stable.

PacBio Revio dramatically increased HiFi output per run and brought the per-base cost of highly accurate long reads down to something competitive for whole genomes. The instrument itself is a serious capital investment, which pushes most labs toward a core facility rather than owning one.

ONT's cost structure is unique because of the hardware tiers. A MinION flow cell is cheap enough to run a small project on a laptop-sized device, while the PromethION scales to population-level throughput. That low barrier to entry, plus real-time streaming basecalls and in-the-field portability, is a category PacBio simply does not compete in. For a small lab, a pathogen outbreak in a clinic, or a remote sampling site, ONT is frequently the only practical option.

Beyond the Genome: Methylation and Direct RNA

Both platforms read DNA modifications natively, without the bisulfite conversion that older methods required, which is a genuine advantage over short reads for epigenetics work. If you are coming from that world, our primer on DNA methylation and bisulfite sequencing explains what native long-read modification calling replaces.

Dorado can emit 5mC and 5hmC calls directly from the same squiggle used for basecalling, and ONT can also sequence native RNA strands directly — no reverse transcription, so you keep RNA base modifications and full-length isoform structure. PacBio offers full-length transcript sequencing (Iso-Seq) with HiFi accuracy, which is excellent for isoform discovery even though it goes through cDNA. For phased, allele-specific methylation across a whole genome, ONT's combination of long reads plus native modification calling is hard to beat.

Downstream Assembly: The Data Shapes the Tool

Your platform choice quietly dictates which assembler you reach for. HiFi's low, random error suits assemblers built to exploit near-perfect long reads, while ONT's longer-but-noisier reads suit tools designed to tolerate indels and stitch across enormous spans. The gold-standard modern approach for many vertebrate genomes actually combines both: HiFi reads for per-base accuracy plus ONT ultra-long reads for scaffolding across the hard repeats.

Whichever route you take, the reads land as FASTQ (see FASTA and FASTQ formats explained) and quickly become alignments in BAM — the same SAM/BAM machinery you already know, just with much longer records and different MAPQ behavior. When we walk through de novo assembly with Flye and hifiasm, the platform-specific flags you pass are a direct consequence of the accuracy and length tradeoffs described here.

A Decision Framework You Can Actually Use

Strip away the benchmarks and the choice usually reduces to a few concrete questions:

• Do you need the cleanest possible per-base calls — for small-variant calling, homopolymer-sensitive gene annotation, or metagenomic strain resolution? Lean PacBio HiFi.
• Do you need to span huge repeats or resolve structural variants and complex loci? Lean ONT ultra-long.
• Do you need portability, real-time results, or the lowest entry cost? Lean ONT, likely MinION or PromethION.
• Do you want a reference-grade, near-complete assembly and have the budget? Use both.

There is rarely a universally correct answer — only the right fit for one project's biology and constraints.

Practical Takeaways

• ONT R10.4.1 with Dorado sup reaches Q20+ modal accuracy for simplex reads (Q30+ with duplex, at lower yield), but its residual errors are still indels in homopolymers — a real concern for annotation and small-indel calling.
• PacBio HiFi from Revio delivers Q30+ reads with rare, substitution-like errors, making it the safer choice when per-base correctness matters most.
• Read length is ONT's decisive edge: ultra-long reads past 100 kb bridge repeats that HiFi's ~15 to 25 kb inserts cannot.
• Budget for GPUs on any ONT project — Dorado sup basecalling is compute-intensive, while HiFi consensus happens on-instrument.
• Both platforms call 5mC/5hmC natively, and ONT uniquely offers direct native-RNA sequencing.
• For reference-grade genomes, combining HiFi accuracy with ONT ultra-long scaffolding is the current gold standard, not an either/or.

Once you have chosen a platform and your reads are in hand, the next step is turning them into contigs. Head to our hands-on walkthrough of de novo genome assembly with Flye and hifiasm to see how these accuracy and length tradeoffs translate into the actual commands and flags you will run.