Apple’s A20 Pro chip has posted Geekbench scores that beat every desktop CPU on the market today, including AMD’s Ryzen 9 9950X3D and Intel’s Core i9-14900KS, in single-core performance. The leaked results, first surfaced in Geekbench Browser entries and confirmed across multiple outlets between September 10 and September 14, 2026, mark the first time a smartphone chip has out-scored the fastest desktop processors money can buy in single-thread work. Inven Global put it bluntly in its September 14 report: “iPhone’s Brain A20 Pro Beats Desktop CPUs and M5 in Single-Core Performance.”
The chip, built on a 2nm-class process, was introduced by Apple on September 9, 2026 with the iPhone 18 Pro and iPhone 18 Pro Max. The new iPhone Duo also uses it. The benchmark figures, however, are third-party claims and not Apple-confirmed results, but their spread across Taiwanese, Japanese, Polish, and US outlets has made this a closely watched hardware story. If the numbers hold up in shipping hardware, Apple will have done something no smartphone maker has managed before: put a phone chip ahead of the best that AMD and Intel can build for a desktop tower.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
The numbers: what A20 Pro actually scored
Two separate benchmark suites are circulating, and it matters which one is being cited. In Geekbench 6, multiple leaked entries put the A20 Pro at 4,719 single-core and 12,677 multi-core, according to Japanese outlet GIGAZINE’s September 11 report and corroborated by Taiwanese site TechNews via LINE Today the same day, with PhoneArena landing on the identical 4,719/12,677 pair in its own September 2026 write-up. Beebom’s reading of the same leak came in marginally lower at 4,707/12,598, a spread of well under half a percent. A second device entry, tied to an iPhone 18 Pro Max test unit, scored 4,725 single-core and 12,575 multi-core; Notebookcheck logged that same 4,725 single-core figure specifically under Geekbench 6.7.1, alongside roughly 12,600 multi-thread points. Tom’s Hardware, which reviewed the same Geekbench Browser submissions, reported the identical 4,719/12,677 split and noted the chip runs at clocks “nearly 5 GHz,” with the Geekbench metadata showing a peak of 4.93 GHz.
A separate wave of scores comes from Geekbench 7, the benchmark suite Inven Global and Tom’s Hardware used for their desktop-CPU comparison stories. There, A20 Pro posted a single-core score between 4,006 and 4,042, with Tom’s Hardware recording 4,006 single-core against 11,460 multi-core and Inven Global citing the higher 4,042 single-core figure in September 2026. Spanish-language outlet NeoTeo pinned the upload date of that Geekbench 7 entry precisely, to September 10, 2026. Tom’s Hardware’s headline framed it directly: the “2nm chip beats desktop Intel Core i9 and AMD Ryzen 9 by up to 32 percent.” Inven Global’s version of the comparison had A20 Pro’s Geekbench 7 score beating the Ryzen 9 9950X3D by roughly 26% and the Intel Core i9-14900KS by around 32%.
Geekbench 6 and Geekbench 7 are not directly comparable benchmarks, so readers should treat the two data sets separately rather than averaging them. The gap is obvious once you line up the multi-core columns: the Geekbench 6 leaks cluster around 12,600 to 12,677, while the Geekbench 7 run Tom’s Hardware documented in September 2026 returned 11,460, a difference that reflects scoring baselines rather than any change in silicon. What’s consistent across every outlet, though, is the direction of the story: on single-threaded work, a phone chip is outscoring the best desktop silicon AMD and Intel currently sell.
| Chip | Benchmark | Single-core score | Source |
|---|---|---|---|
| Apple A20 Pro | Geekbench 6 | 4,719 (also 4,725) | GIGAZINE, TechNews/LINE Today, Tom’s Hardware |
| Apple A20 Pro | Geekbench 7 | 4,006–4,042 | Inven Global, Tom’s Hardware |
| Apple M5 (MacBook Pro) | Geekbench 6 | 4,263 | MacRumors, Cult of Mac, Tom’s Guide |
| AMD Ryzen 9 9950X3D | Geekbench 7 | ~26% behind A20 Pro | Inven Global |
| Intel Core i9-14900KS | Geekbench 7 | ~32% behind A20 Pro | Inven Global, Tom’s Hardware |
| Apple A19 Pro (prior gen) | Geekbench 6/7 | ~23–25% behind A20 Pro | Tom’s Hardware, GIGAZINE |
A20 Pro vs. Apple’s own M5: an awkward internal race
The more surprising part of this story isn’t that a phone chip is beating x86 desktop silicon, it’s that A20 Pro appears to be beating Apple’s own laptop and desktop chip too. The M5, which launched inside the 14-inch MacBook Pro in late 2025, set what MacRumors and Cult of Mac both called the highest single-core Geekbench 6 score “ever recorded in the Geekbench 6 database for any Mac or PC processor” at the time, with a leaked score of 4,263. That record stood for roughly eleven months.
Wccftech’s September 10 coverage described the A20 Pro’s 4,719 Geekbench 6 result as beating “M5 Max,” the higher-tier variant of the M5 family, and called it a new record for the fastest single-core CPU score in the Geekbench 6 database, period, not just for phones. Poland’s PurePC ran a nearly identical story on the same day, also crediting the A20 Pro with breaking the single-core record and surpassing M5 Max.
That creates an odd product-line dynamic for Apple. The company has spent the past five years positioning its M-series Mac chips as the performance flagship of its silicon lineup, with A-series phone chips treated as a smaller, power-constrained sibling. If A20 Pro’s leaked numbers hold in production hardware, Apple’s cheapest current iPhone-class silicon will out-score the chip inside its $1,599-and-up MacBook Pro line on single-thread work, at least until M6 arrives.
Why single-core still matters in 2026
Multi-core benchmarks get more headlines because they scale with core count and look good in marketing slides, but single-core performance is what most everyday software actually leans on. Web browser tab switching, UI responsiveness, single-threaded game logic, and a large share of Adobe’s creative suite still bottleneck on the speed of one core doing sequential work as fast as possible. A chip that wins single-core by a wide margin tends to feel faster in daily use even when its total multi-core throughput trails a workstation-class part with far more cores.
That’s part of why Apple’s single-core lead has become a recurring storyline since the M1 launched in 2020. Each generation of Apple Silicon has either held or reclaimed the single-core crown against x86 rivals, largely on the back of aggressive per-core design choices, wide execution pipelines, and TSMC’s leading-edge process nodes, rather than simply adding more cores. The A20 Pro leak, if accurate, extends that pattern from laptops and desktops into phones for the first time.
The 2nm process node: what’s actually new
Every outlet tracking the leak agrees on one manufacturing detail: A20 Pro is built on a 2nm-class process, widely understood in the industry to mean TSMC’s N2 node, Apple’s foundry partner for every A-series and M-series chip to date. Tom’s Hardware’s September 10 report called it “the 2nm CPU in the iPhone Duo and 18 Pro,” while TechNews’ Taiwanese coverage cited the same 2nm classification tied to the same Geekbench 6 entries.
Moving from N3 (used in A18 Pro and A19 Pro) to N2 is TSMC’s most significant node transition since the jump to 3nm, promising meaningful gains in transistor density and power efficiency. That’s consistent with the leaked generational uplift: Tom’s Hardware calculated roughly a 23.3% single-thread and 27.1% multi-thread improvement over A19 Pro from the Geekbench 7 data, while the Geekbench 6 leaks point to a similar ~25% uplift across both single- and multi-core scores. GIGAZINE’s write-up echoed the same generational delta.
The leaked Geekbench metadata listed a 6-core CPU configuration, without a public performance/efficiency core breakdown. Reporting on the released chip describes a six-core CPU, a seven-core GPU, and a dual 16-core Neural Engine. Given Apple’s design pattern across the last several A-series generations, a split weighted toward efficiency cores is likely, but that specific ratio has not been confirmed by Apple or by any of the outlets reporting the leak. One report also says it has increased memory bandwidth relative to A19 Pro.
Which iPhones will actually ship with it
The Geekbench entries tie A20 Pro to test devices identified as iPhone19,2 and a separate iPhone 18 Pro Max unit, both running iOS 27.0 pre-release builds. GIGAZINE and Tom’s Hardware both describe the chip as destined for the iPhone 18 Pro, iPhone 18 Pro Max, and a device line referred to as the iPhone Duo. Apple introduced the iPhone 18 Pro and iPhone 18 Pro Max with the A20 Pro in September 2026, and the iPhone Duo is reportedly a dual-screen or dual-form-factor device that uses the same chip. The chip supersedes A19 Pro in the reported iPhone Pro lineup, as a generation replacement.
Apple introduced the A20 Pro on September 9, 2026, so the launch date is known. The benchmark coverage, by contrast, comes from third-party leaks that mention pre-release iOS 27.0 test builds, and none of it is an Apple-published result, so the scores should be read as unconfirmed claims.
How reliable are leaked Geekbench scores, really?
Every outlet covering this story, without exception, flagged the same caveat: these are unconfirmed, early submissions to the public Geekbench Browser database, not numbers Apple has published or endorsed. Tom’s Hardware’s own language on the Geekbench 7 scores was direct, describing them as coming from “an early submission (which may or may not demonstrate performance of actual A20 Pro-based products, so take the results with a grain of salt).” MacRumors used nearly identical hedging language when covering the M5’s leaked Geekbench 6 score a year earlier, calling it “based on a single unconfirmed result uploaded to the Geekbench 6 database,” a caution Tom’s Guide echoed in its own coverage of the same leak.
There are legitimate reasons for skepticism beyond boilerplate caution. Pre-production engineering samples sometimes run at non-representative clock speeds, either throttled for validation testing or, less commonly, overclocked beyond what shipping silicon will sustain. Geekbench device identifiers can in theory be mislabeled, though major outlets generally filter out obvious spoofing before publishing. And comparing Geekbench 6 to Geekbench 7 results, as some casual social coverage has done, produces numbers that look more dramatic than they are because the two benchmark versions use different scoring baselines.
None of that means the leak is wrong. Multiple independent outlets working from separate Geekbench Browser entries landed on consistent figures: 4,719 single-core in Geekbench 6, a peak clock near 4.93 GHz, a 6-core CPU, and a 2nm process. PhoneArena reproduced the 4,719/12,677 pair exactly, Beebom came within 12 points at 4,707/12,598, and Notebookcheck’s Geekbench 6.7.1 reading of 4,725 and roughly 12,600 multi-thread points brackets the same range from the other side, all in September 2026. That kind of cross-outlet consistency is generally treated as high-confidence in the smartphone leak ecosystem, even without Apple’s sign-off.
What this means for AMD and Intel
AMD and Intel aren’t losing customers to iPhones. Nobody is replacing a gaming rig or a workstation with a smartphone. But the optics matter for a different reason: single-core leadership has become the industry’s shorthand for who has the best CPU microarchitecture, independent of how many cores a chip can cram onto a die. When a phone SoC constrained by a battery and a fanless enclosure outscores a desktop chip drawing 200+ watts under an air or liquid cooler, it says something uncomfortable about the state of x86 per-core efficiency.
This isn’t a new problem for AMD and Intel. Apple Silicon has led or matched x86 in single-core efficiency-adjusted performance since the M1 in 2020, and the M5’s 4,263 single-core score already had Tech Insider’s own May 2026 comparison showing the M5 beating Intel’s Core Ultra 9 285K by roughly 31% in single-threaded work. What’s new with A20 Pro is that the lead now extends into a category, phone chips, where nobody expected desktop-class comparisons to even be relevant.
Intel’s roadmap response is already partly public: the company has signaled CPU price increases of roughly 10% ahead of its next-generation desktop launch expected around March 2027, a move some analysts read as an attempt to fund more aggressive node and architecture investment rather than simply passing along cost pressure. AMD, meanwhile, continues to lean on 3D V-Cache and chiplet designs like the Ryzen 9950X3D to compete on multi-threaded and gaming workloads where core count and cache still dominate over pure single-thread clock speed.
Historical context: Apple’s single-core streak
Apple’s single-core dominance story didn’t start with A20 Pro. The M1, launched in November 2020, was the first Apple Silicon chip to post desktop-competitive single-core Geekbench scores from a fanless design, catching the PC industry off guard. Every subsequent M-series and flagship A-series generation has either extended or defended that lead, moving through TSMC’s 5nm, 4nm, and 3nm nodes on the way to the 2nm node A20 Pro is reportedly using now.
The pattern accelerated in 2025 and 2026. The M5’s 4,263 single-core Geekbench 6 score, leaked in October 2025, was described by MacRumors as the highest ever recorded in the Geekbench 6 database for any Mac or PC processor at that point. Less than a year later, A20 Pro’s leaked 4,719 reportedly broke that same record from inside a phone. That’s a fast internal handoff, and it underscores how much of Apple’s competitive advantage now comes from process-node access and core-design discipline rather than raw power budget.
| Generation | Chip | Node | Year | Notable claim |
|---|---|---|---|---|
| 1 | Apple M1 | TSMC N5 (5nm) | 2020 | First fanless chip to match desktop single-core scores |
| 4 | Apple A19 Pro | TSMC N3 | 2025 | Prior-gen iPhone flagship baseline |
| 5 | Apple M5 | TSMC N3/next-gen | 2025 | 4,263 single-core, record at the time per MacRumors |
| 6 | Apple A20 Pro | 2nm-class | 2026 | Introduced with the iPhone 18 Pro and iPhone 18 Pro Max |
Competitive comparison: phone chip vs. laptop chip vs. desktop chip
It’s worth separating three categories that are getting conflated in social media coverage of this leak. First, A20 Pro is a smartphone SoC, thermally constrained by a phone chassis with no active cooling. Second, M5 is a laptop and desktop-class chip with a larger thermal envelope and, in Mac Studio or Mac Pro configurations, active cooling and sustained power delivery. Third, Ryzen 9950X3D and Core i9-14900KS are unconstrained desktop chips with 200W-plus power budgets and dedicated cooling solutions.
That A20 Pro can reportedly out-score both categories on single-core, despite having by far the smallest power and thermal budget of the three, is the actual headline here, not simply “phone beats desktop.” A20 Pro doesn’t out-multi-thread a 16-core desktop chip, and none of the current reporting claims it does. It wins narrowly on the specific metric of how fast one core can execute sequential instructions, which is a proxy for microarchitectural efficiency rather than total compute throughput.
Market impact: what changes for buyers this fall
For the average iPhone buyer, a Geekbench leak doesn’t change much in the short term, since real-world responsiveness gains from a faster chip tend to be incremental year over year rather than transformative. But the leak does feed the marketing narrative around A20 Pro: rather than framing it purely against last year’s A19 Pro, coverage invites direct comparisons with PC and Mac hardware, a framing Apple has used sparingly for iPhone chips in the past but leans on heavily for M-series marketing.
For PC and Mac buyers eyeing an upgrade this fall, the more relevant signal is what A20 Pro implies about the next M-series chip. If Apple’s phone silicon is already matching or beating the current M5 generation on single-core, the M6, expected to follow the company’s roughly annual M-series cadence, will likely need a larger architectural jump than usual to maintain separation between Apple’s phone and Mac product lines.
Predictions: where this goes next
Based on the pattern of prior Apple Silicon launches and the current state of the leak, a few outcomes look likely now that A20 Pro has been introduced:
- Apple introduced A20 Pro on September 9, 2026, and independent single-core Geekbench numbers will likely land close to, but not necessarily identical to, the leaked 4,719 figure, consistent with how prior A-series and M-series launches have tracked their pre-release leaks.
- Watch whether Apple’s marketing materials explicitly reference desktop CPU comparisons for the first time in an iPhone-chip context, which would be a shift from its historical practice of comparing new A-series chips mainly to their direct predecessor.
- AMD and Intel are unlikely to respond with immediate architecture changes, since neither company treats phone SoCs as direct competition, but expect both companies’ marketing teams to emphasize multi-core, gaming, and workstation throughput metrics where x86 still holds a clear advantage over any phone or laptop chip.
- The M6, whenever it arrives, will need to clear a higher single-core bar than M5 did to avoid the awkward optics of a Mac chip trailing an iPhone chip on Apple’s own generation-over-generation slides.
- Expect a wave of “is your $2,000 PC slower than an iPhone” style coverage across tech YouTube and enthusiast forums now that A20 Pro has shipped in the iPhone 18 Pro and iPhone 18 Pro Max and independent reviewers can run controlled, non-leaked benchmarks.
What to watch for after Apple’s official announcement
The clearest confirmation signal is Apple’s own published material, where the company traditionally discloses core counts, process node, and relative performance claims against the prior generation. Independent reviewers with retail hardware, rather than pre-release engineering samples, will also be the real test of whether the leaked 4,719 single-core figure holds up outside a controlled leak environment. The first data point on that front has already arrived: in October 2026, PhoneArena measured a shipping iPhone 18 Pro Max at 4,047 single-core and 11,648 multi-core, landing just above the 4,006-to-4,042 Geekbench 7 window the September leaks had sketched out. Historically, Apple’s own on-stage performance claims have landed close to, though sometimes slightly more conservative than, the boldest pre-launch leaks.
It’s also worth watching whether Apple addresses the M5-vs-A20-Pro overlap directly. The company has generally avoided drawing attention to internal performance overlaps between its phone and Mac chip lines, preferring to let each product’s marketing stand on its own. A leak this widely reported may force the question at a press briefing, even if Apple doesn’t address it from the main stage.
Frequently asked questions
Is the A20 Pro benchmark officially confirmed by Apple?
No. Every outlet reporting the scores, including Tom’s Hardware, GIGAZINE, TechNews, PhoneArena, Beebom, and Notebookcheck, describes them as leaked, unconfirmed Geekbench Browser submissions, not official Apple specifications. NeoTeo dated the original Geekbench 7 upload to September 10, 2026, which is the earliest fixed point in the timeline rather than an Apple disclosure.
What Geekbench 6 single-core score did A20 Pro reportedly achieve?
Leaked entries show 4,719 and a separate 4,725 result, according to GIGAZINE, TechNews/LINE Today, and Tom’s Hardware. PhoneArena reported the same 4,719/12,677 single/multi-core pair in September 2026, Beebom logged 4,707/12,598, and Notebookcheck recorded 4,725 under Geekbench 6.7.1 with roughly 12,600 multi-thread points.
Does A20 Pro really beat the Ryzen 9 9950X3D and Core i9-14900KS?
In Geekbench 7 single-core testing, Inven Global (4,042) and Tom’s Hardware (4,006 single-core, 11,460 multi-core) both reported A20 Pro’s leaked score of roughly 4,006-4,042 beating those desktop chips by approximately 26-32%. Those September 2026 figures were unconfirmed early submissions rather than independently verified retail comparisons, though PhoneArena’s October 2026 measurement of a retail iPhone 18 Pro Max at 4,047/11,648 tracks closely with them.
Which iPhones will use the A20 Pro chip?
Apple introduced A20 Pro with the iPhone 18 Pro and iPhone 18 Pro Max on September 9, 2026. Reports from GIGAZINE and Tom’s Hardware also tie it to a device referred to as the iPhone Duo.
What manufacturing process does A20 Pro use?
Multiple outlets, including Tom’s Hardware and TechNews, describe it as a 2nm-class chip, understood in the industry to mean TSMC’s N2 process node.
How does A20 Pro compare to Apple’s own M5 chip?
Wccftech and PurePC both reported that A20 Pro’s leaked Geekbench 6 single-core score of 4,719 beats the M5 Max variant, which had previously held the record with a score of 4,263 according to MacRumors and Cult of Mac.
How much faster is A20 Pro than the A19 Pro?
Tom’s Hardware calculated roughly a 23.3% single-thread and 27.1% multi-thread improvement over A19 Pro based on the leaked Geekbench 7 scores, while separate Geekbench 6 leak coverage points to a similar ~25% generational uplift.
Why do leaked benchmark scores sometimes turn out to be wrong?
Pre-production engineering samples can run at non-representative clock speeds, and single isolated database entries may not reflect final shipping silicon. Every outlet covering this leak, including Tom’s Hardware and MacRumors, explicitly cautioned that the figures are unconfirmed.
