On Monday, a Hacker News post titled “Xiaomi: New CPU matches Apple cores single threaded, much faster multithreaded” pulled in 716 points and 478 comments. The chip in question is the XRing O1, Xiaomi’s first in-house smartphone SoC. The word “new” is doing a lot of work: the XRing O1 was announced on May 22, 2025 — fifteen months ago.

The numbers don’t support the headline either. Early Geekbench results for the XRing O1 show 2,723 single-core and 8,665 multi-core. Apple’s A18 Pro — the chip the post claims Xiaomi “matches” — scores 3,358 single-core. That’s a 19% gap, not a match. Multithreaded, the XRing O1’s 8,665 beats the A18 Pro’s 8,184 by about 6% — not “much faster,” and still behind the Snapdragon 8 Elite’s 9,271.

So a fifteen-month-old chip with middling benchmark scores went viral this week because someone posted a tweet with an inflated claim, and 716 people upvoted it without checking the numbers. That’s the story. Not the chip. The story is the story.

The Numbers Were Always There

HotHardware published the XRing O1’s first Geekbench cameo back in May 2025, and the numbers were unremarkable then. The chip slots in around the Dimensity 9400, behind the Snapdragon 8 Elite, and well behind Apple’s A18 Pro in single-threaded performance. Notebookcheck’s review reached the same conclusion: the XRing O1 is “slightly above a Dimensity 9400 in the top class, but has to admit defeat to the Snapdragon 8 Elite.”

None of this was hidden. The data was public for over a year. What changed this week wasn’t the chip — it was the packaging. A tweet with a confident claim, a HN title that amplified it, and suddenly a 19% single-core deficit became “matches Apple.” The 478 comments are debating a claim that doesn’t match the data, and the data has been sitting in public view since May 2025.

Geekbench 7 Changed the Rules in July

Primate Labs released Geekbench 7 on July 23, 2026, adding AI workloads and redesigning the benchmark to “better reflect 2026 computer use cases,” as AppleInsider reported. The timing matters. The new benchmark has re-opened the leaderboard, and every chipmaker’s scores are being re-litigated. The XRing O1’s viral moment is the first big test of how Geekbench 7 numbers get laundered through social media.

The pattern is familiar to anyone who’s watched a benchmark cycle before: a new version drops, old scores get re-run, and the discourse fills with claims that outrun the data. But Geekbench 7’s AI workloads add a new wrinkle. The benchmark now measures things that are harder to verify independently, which means the gap between the claim and the data is wider than ever. A number that used to mean something specific now means whatever the poster says it means.

The Benchmark Is the Product Now

Here’s the uncomfortable part. The benchmark-industrial complex used to serve engineers. Primate Labs built Geekbench so hardware reviewers could compare chips on a common scale. Now the benchmark serves a different customer: the content feed. The number isn’t the product. The post about the number is the product.

A junior engineer at a competing chipmaker put it plainly in a Slack DM this week, passing around the HN thread: “Nobody in that thread has run the benchmark. They’re arguing about a tweet about a benchmark.” He’s right. The 478 comments are debating a claim that doesn’t match the data, and the data has been public for fifteen months.

This isn’t a Xiaomi problem. It’s a discourse problem. The benchmark was supposed to be the objective referee — the thing that settles arguments with numbers. Instead, it’s become the raw material for arguments that don’t need numbers at all. The claim matters more than the data, and the data is just a prop.

The XRing O1 is a perfectly fine first effort from Xiaomi. It’s competitive with MediaTek, behind Qualcomm, and nowhere near Apple in single-threaded performance. That’s a reasonable outcome for a first-generation in-house SoC. But “reasonable outcome” doesn’t go viral. “Matches Apple” does.

So the next time you see a benchmark post with 700 upvotes and a confident headline, ask yourself: is this a number, or is this a story about a number? Because in 2026, those are different things.

Sources