It looks like the first benchmarks of Apple’s upcoming M2 Max chip have leaked in Geekbench’s database.

When users run the over-the-shelf version of the Geekbench 5 benchmarking tool, the scores are logged to a public database of results and are tied to entries for specific hardware. In this case, the result (which was discovered by a Twitter user) is listed under a product labeled “Mac14,6” running the as-yet-unreleased operating system “macOS 13.2 (Build 22D21).” The entry also noted that the chip had 12 cores.

The chip in question is likely destined for MacBook Pro and Mac Studio models that will launch sometime next year. As for the results: The overall single-core score is 1,853, and the multicore score is 13,855. The more granular scores like crypto, integer, and floating point generally track along the same lines when compared to this chip’s predecessor, the M1 Max.

If those results are both legit and representative of the final product, they might be a disappointment for some. Compared to our Geekbench 5 tests of a 16-inch MacBook Pro with the M1 Max, that’s around a 6 percent improvement in single-core performance and an 11 percent bump for multicore.

That performance difference isn’t too far off from comparisons between the lower-end M1 and M2 that have already been introduced, though, so it’s not too surprising if it’s accurate.

That said, there are several caveats to consider. First, these synthetic benchmarks can be good for a ballpark idea of relative performance between devices, but that doesn’t mean they always reflect real-world performance. Second, the M2 Max results seen here might not represent the final product. Finally, they might not be real at all.

So yes, those are a lot of caveats. Ars readers will know to take this with an entire pound of salt—but it’s interesting to see and speculate about in any case.

https://arstechnica.com/?p=1901150