Skip to content
News and Updates

Apple Unveils Mac Mini M6 and M5 Pro: Built for On-Device AI

Jared Whitman
Aug 26, 2026  /  7 min read
Apple Mac mini M6 held in a hand, showing its compact aluminum enclosure and front ports
Image: Apple.

Apple’s new Mac mini starts at $899, ships with a chip that’s 4x faster at AI workloads than the outgoing M4, and goes on sale September 22, 2026 — but the headline isn’t the specs, it’s that Apple built its cheapest desktop around running AI models locally instead of leaning on the cloud. The company announced the refreshed Mac mini, powered by the new M6 and M5 Pro chips, on August 25, 2026, alongside an updated Mac Studio, and buried in the marketing copy is a fairly blunt admission: Apple now sees “on-device AI” as the reason to buy a $900 box, not “it’s small and quiet.”

I’ve bought three Mac minis over the past decade, mostly as cheap always-on servers tucked behind a monitor. None of them were ever pitched to me as an AI machine. This one is, explicitly, and that says something about where the whole industry thinks computing is headed in late 2026.

What’s actually new in the M6 and M5 Pro

The base Mac mini now runs on the M6: a 12-core CPU, a 12-core GPU with Neural Accelerators built into every GPU core, and a new dual 16-core Neural Engine. Apple’s own numbers claim 4x faster AI performance than the M4 model it replaces, 2x faster graphics and storage throughput, and a 40% bump in raw CPU performance. Unified memory starts at 16GB and can be configured up to 32GB, with 170GB/s of memory bandwidth feeding it.

Step up to the M5 Pro model and you get up to an 18-core CPU and a 20-core GPU with third-generation ray tracing, up to 64GB of unified memory, and 307GB/s of bandwidth — nearly double the base chip. Apple says the M5 Pro handles large language model processing 4x faster than the M4 Pro it replaces, and renders about 1.4x faster in Blender, a benchmark Apple rarely cites unless it thinks the number is genuinely good.

Connectivity got a quiet but meaningful refresh too: both chips move to Wi-Fi 7 and Bluetooth 6, with 2.5Gb Ethernet standard (10Gb available as an option). The base M6 mini keeps three Thunderbolt 4 ports, while the M5 Pro jumps to three Thunderbolt 5 ports. Both keep the two front-facing USB-C ports, HDMI, and a headphone jack that Apple, mercifully, still hasn’t removed from its desktops.

Quick-reference: Mac mini M6 vs. M5 Pro

SpecMac mini (M6)Mac mini (M5 Pro)
CPU12-coreUp to 18-core
GPU12-core w/ Neural AcceleratorsUp to 20-core, 3rd-gen ray tracing
Unified memory16GB (up to 32GB)Up to 64GB
Memory bandwidth170GB/s307GB/s
AI perf vs. predecessor4x vs. M44x LLM processing vs. M4 Pro
Thunderbolt3x Thunderbolt 43x Thunderbolt 5
Starting price$899 ($799 edu)$1,699 ($1,599 edu)
ShipsSeptember 22, 2026 (pre-orders open now)

Why Apple is pitching a mini desktop as an AI machine

Apple’s hardware chief Johny Srouji put it plainly in the announcement: the Mac mini with M6 “delivers a whole new level of AI performance.” That’s not an accident of marketing — it’s the second half of a strategy Apple has been building toward since Apple Intelligence launched, and it’s the same story Apple told with its delayed smart glasses: the company keeps choosing to hold products back or reposition them around doing more processing on the device itself, rather than routing everything through a data center.

There’s a practical reason for that beyond privacy marketing. Running a large language model locally means no network round-trip, no per-query cloud cost, and no dependence on a subscription tier to get useful agentic behavior out of your machine. Apple explicitly frames the new Mac mini as hardware for “running local AI models and powering always-on agentic computing tasks” — language that reads like it’s aimed squarely at developers and power users experimenting with local models, not just people editing photos.

It also puts Apple in a slightly different lane than the rest of the industry’s AI hardware race. While companies like Nvidia and Cerebras are racing to build bigger inference clusters for data centers, and chipmakers are locking in multi-billion-dollar supply deals — see Google’s recent $12.2 billion Marvell chip stake — Apple’s answer is a compact, no-fan desktop under $900 that can run a meaningful model without touching the cloud at all. It’s a much smaller bet, but it’s aimed at a market — individual developers, small studios, tinkerers — that the giant data-center deals don’t really serve.

The Mac Studio got the same treatment

Apple didn’t just refresh the mini. The same announcement introduced an updated Mac Studio built around the new M5 Max and M5 Ultra chips, continuing the same “more local AI horsepower” theme at the high end of the desktop lineup. Apple hasn’t published full Mac Studio specs as prominently as the Mac mini’s, but the timing — both machines refreshed on the same day, both leaning on the same AI-performance language — makes it clear this is a coordinated repositioning of the entire desktop line around on-device AI, not just a spec bump for the cheapest model.

It’s also a notable move given the memory market Apple is buying into right now. RAM and NAND prices have been climbing for over a year as AI data centers consume a growing share of global output, and Apple is still shipping a 16GB base configuration on its cheapest machine rather than cutting it further to hit a price point — a decision that will matter more the longer the memory shortage runs.

Should you actually buy one

If you already own an M4 Mac mini and mostly use it for everyday tasks — browsing, office work, media playback — this isn’t a must-upgrade. The 40% CPU gain and 2x graphics bump are real, but you won’t feel them doing email. Where this refresh actually matters is if you want to run local AI models: coding assistants, image generation, or agentic tools that benefit from more Neural Engine throughput and more unified memory to hold a bigger model in RAM. For that specific use case, the jump from 16GB to a configurable 32GB (or up to 64GB on the M5 Pro) is the number that counts, more than the CPU or GPU core totals.

For most other buyers, the base M6 Mac mini at $899 is still one of the more sensible ways into the Mac ecosystem — it’s just no longer being sold to you as a cheap desktop that happens to run macOS. It’s being sold as a small AI appliance that also happens to run macOS.

FAQ

When does the new Mac mini go on sale?

Apple opened pre-orders on August 25, 2026, with the Mac mini M6 and Mac mini M5 Pro both shipping starting September 22, 2026.

How much does the new Mac mini cost?

The M6 model starts at $899 ($799 for education pricing). The M5 Pro model starts at $1,699 ($1,599 for education).

Is the new Mac mini actually faster at AI tasks, or is that just marketing?

Apple’s published numbers claim 4x faster AI performance on the M6 versus the outgoing M4, and 4x faster large language model processing on the M5 Pro versus the M4 Pro. Those are Apple’s own benchmarks, not independently verified yet, but the underlying hardware change — Neural Accelerators built into every GPU core plus a larger dual 16-core Neural Engine — is a real architectural shift, not just a clock-speed bump.

Does the new Mac mini support more RAM than before?

Yes. The base M6 model now goes up to 32GB of configurable unified memory, and the M5 Pro model supports up to 64GB — both increases that matter specifically for running larger local AI models.

Did Apple update anything else alongside the Mac mini?

Yes, Apple announced an updated Mac Studio on the same day, built around new M5 Max and M5 Ultra chips, continuing the same on-device AI performance push across its desktop lineup.

Written by
Jared Whitman

Jared is a tech journalist covering product launches, industry news, and the culture around technology. He has been reporting on the consumer tech beat for more than eight years.

Up Next