# Guessing the Performance of M5 Pro Recent news claims that M5 MacBooks have been [_nearing_](https://9to5mac.com/2025/09/28/new-macbook-air-macbook-pro-and-studio-display-nearing-mass-production-report/) the "mass production" phase. This can help us gain some insights on the release time of the new MacBook models: In 2024, mass production started in August and the devices were released in October. So, drawing parallels to this year, if mass production starts in October, the devices should be released in December, which can practically January, matching the [rumors](https://appleinsider.com/articles/25/09/28/new-apple-displays-m5-macbook-upgrades-expected-by-q1-2026). But putting the "release time" aside: in this post, I try to guess the real technical differences between M4 Pro and M5 Pro. For simplicity, we assume that the M-series processors of Mac are kind of scaled-up versions of the A-series processors of iPhone. M4 Pro can be considered as the "bumped up" version of A18 Pro on iPhone 16 Pro, so M5 Pro could be considered bumped up from A19 Pro on iPhone 17 Pro and iPhone Air. As a result, observing the jump from A18 Pro to A19 Pro might give us a rough estimate of differences between M4 Pro and M5 Pro. Particularly, with the transform from A18 Pro to A19 Pro, in the benchmarks, we saw: - 9% higher single-core and 10% higher multi-core CPU score ([link](https://nanoreview.net/en/soc-compare/apple-a19-pro-vs-apple-a18-pro)) - 40% faster GPU ([link](https://www.notebookcheck.net/Apple-A19-Pro-hands-on-GPU-benchmarks-show-40-upgrade-versus-A18-Pro-but-the-Snapdragon-8-Elite-is-hardly-inferior.1117278.0.html)) - 26% higher memory bandwidth ([link](https://nanoreview.net/en/soc-compare/apple-a19-pro-vs-apple-a18-pro)) - "Neural accelerator" inside each GPU core, for faster inference of local LLMs - One [user](https://www.reddit.com/r/LocalLLaMA/comments/1nlu3cd/the_iphone_17_pro_can_run_llms_fast/) claimed 10x faster prompt processing and 2x faster token generation; another [user](https://x.com/JeffreyLind2/status/1969904651190387084) claimed 10-30% more tokens/second - A [video](https://x.com/adrgrondin/status/1969484236919869931) showing how fast the inference is on A19 Pro - Possibly better battery and sustained performance (can't be quantified because the battery size and thermal design of iPhones has also changed over time) I expect similar gains for the transition between M4 Pro and M5 Pro as well. Of course, in reality, things might not be this simple (for example, Apple can choose to use a different combination of performance and efficiency cores, optimizing either battery life or overall performance), but this might be a rough estimate. Let's wait for official benchmarks to see if this guess holds true!