New Mac mini quietly appears on the official website, possibly the most affordable local AI host. Apple has always had a habit of holding a launch event for every major product. But not this time. On August 25, Apple's official website directly listed the new Mac mini—no teaser, no invitation, no Cook's keynote; it was just there when you opened the site. The M6 version starts at 6,999 RMB in China.

💡 What You Will Learn

New Mac mini quietly appears on the official website, possibly the most affordable local AI host. Apple has always had a habit of holding a launch event for every major product. But not this time. On

📜 Table of Contents

• The first 2nm chip (skipping the M5 is a rare move)
• Compared to the M4 machine: how big is the real jump?
• Why it's being called a local AI host
• Don't forget the Mac Studio that launched alongside it (the real performance monster)
• But on price, let me pour some cold water first

New Mac mini quietly appears on Apple's website — possibly the cheapest local AI host

Apple has always had a habit of holding a launch event for every major product. But not this time.

On August 25, Apple's website simply listed the new Mac mini — no teaser, no invitation, no Cook keynote. Open the site and there it is. The M6 version starts at ¥6,999 in China (16GB+256GB), with pre-orders opening at 9 AM on August 27 and official sales starting September 22.

The first 2nm chip (skipping the M5 is a rare move)

The star here is the M6 chip (the latest generation of Apple's in-house computer silicon). It has two "firsts": Apple's first M-series chip built on a 2nm process, and the first time the Mac mini has skipped an entire generation — the previous model ran the M4, and the base M5 was simply passed over.

The M6 packs a 12-core CPU and 12-core GPU, with Apple claiming it delivers "the world's fastest single-threaded performance." The CPU includes two super cores, four performance cores, and six efficiency cores, with multi-threaded performance up to 1.2x that of the M5.

Compared to the M4 machine: how big is the real jump?

I've organized the official data into a table for you:

That 4x AI performance is the headline. For the first time, neural accelerators (small units dedicated to AI computation) have been built into every GPU core, alongside dual 16-core Neural Engines, and memory bandwidth has been boosted from 120GB/s to 170GB/s. Apple's own tests show that in LM Studio (a local LLM runner), prompt processing speeds reach up to 4.8x that of the M4 version.

Why it's being called a local AI host

The direction of this upgrade is crystal clear: local AI.

Wi-Fi 7, Bluetooth 6, and a 2.5Gb Ethernet port are all standard (powered by Apple's in-house N1 wireless chip); the higher-end M5 Pro version also includes Thunderbolt 5 ports — via Thunderbolt 5, multiple Mac minis can be linked into a compute cluster to run large models that exceed a single machine's memory. You could set up a small rack at home and run an open-source LLM that a cloud API wouldn't let you use freely.

Don't forget the Mac Studio that launched alongside it (the real performance monster)

Also newly listed on Apple's site is the updated Mac Studio, featuring the M5 Max and the debut of the M5 Ultra.

The M5 Ultra is the first quad-die 3nm architecture in Apple silicon history — in simple terms, it fuses two M5 Max dies together with 4.4TB/s of interconnect bandwidth. It offers up to a 36-core CPU, 80-core GPU, and 512GB of unified memory, with memory bandwidth of 1.2TB/s — 50% more than the previous M3 Ultra. AI performance reaches up to 4.3x that of the M3 Ultra.

Pricing starts at around ¥18,000 for the M5 Max version and ¥39,000 for the M5 Ultra version, with the fully loaded 512GB configuration shipping in late October. Apple's official positioning for it is "the ultimate desktop device for on-device AI" — one machine can run massive-parameter models locally without kowtowing to cloud API limits.

One mini is the entry point, one Studio is the ceiling — Apple has now covered local AI from entry-level to top-tier.

But on price, let me pour some cold water first

The M4 Mac mini launched at just ¥4,999, and with platform subsidies and national discounts, you could grab it for around ¥3,000. This M6 generation starts at ¥6,999 — a full ¥2,000 more.

Performance has indeed doubled, but the price has also genuinely climbed 40%. I've always advised friends: don't impulse-buy tech on launch day. The M4 version is still available through retail channels and is actually the smarter deal; if you truly want that 2nm AI performance, it's worth waiting for Double 11 to see how prices shake out.

In 2026, does your next computer still need to be a new Mac mini? Sound off in the comments.