Sunk Cost sunkcost.ai Data checked 2026-09-03

Mac mini M4 vs M6, 24GB for local AI

Both hold 13 of the 39 open models here. The Mac mini M4, 24GB costs $300 less. On Gemma 4 12B, the strongest model both hold, the Mac mini M6, 24GB is about 1.5× faster: 16 tok/s against 11, both estimated from memory bandwidth. The Mac mini M4, 24GB pays for itself sooner, in 66 years against 85 years at 500k tokens a day. The Mac mini M4, 24GB is the previous generation, so every figure here for it is priced at what it launched at rather than at a price you can pay today.

Mac mini M4, 24GBMac mini M6, 24GB
Price$799$1,099
Memory24 GB24 GB
Usable by the GPU16 GB16 GB
Memory bandwidth120 GB/s170 GB/s
Power under load65 W65 Wstand-in
Models that fit1313
Best model it runsGemma 4 12BGemma 4 12B
Speed on that model11 tok/s estimated16 tok/s estimated
Pay-back on that modelPays back in 66 yearsPays back in 85 years

Run the numbers on the Mac mini M4, 24GB · or the Mac mini M6, 24GB

What the newer chip buys

These are the same machine a generation apart, in the same case and at the same memory size, so this is the upgrade question rather than a choice between two things on sale. What separates them is bandwidth: the M6 reads its memory at 170 GB/s where the M4 reads it at 120. Decoding reads the whole model out of memory for every token it writes, so that is the figure the speeds in the table follow. The M6 here is listed as 12-core CPU / 12-core GPU, against the M4's 10-core CPU / 10-core GPU.

The two power figures are equal for a reason that is not about either machine: Apple has not published one for the M6, so the data stands the M4's published 65 W in for it and this page prices the electricity into both columns from that. Read it as a placeholder, not as a finding that the newer chip draws the same.

Apple stopped selling the Mac mini M4, 24GB on 25 August 2026, so the $799 above is the price it launched at, and every figure on this page for it is priced at that. A used or refurbished one costs whatever it costs, and pay-back follows the price rather than the machine. Open the calculator on the Mac mini M4, 24GB and put in what you would actually pay; the years move with it. Its own page has what the data records about buying one now.

How much use it takes to pay back

Everything above is at 500k tokens a day. Pay-back moves with how much you actually run, so here are both machines on Gemma 4 12B, the strongest model both hold, at the five levels of use the calculator names. The Mac mini M4, 24GB pays back sooner at every level up to 4M tokens a day. Above that the Mac mini M6, 24GB does.

A day's useMac mini M4, 24GBMac mini M6, 24GB
50ka few chats a day662 years846 years
200klight assistant use166 years212 years
1Ma moderate coding-assistant day33 years42 years
4Mheavy coding with an agent8.3 years11 years
20Magents running most of the day2.1 yearsits ceiling2.1 years

On Gemma 4 12B the Mac mini M4, 24GB generates at most 15.6M tokens a day, so its figure at 20M tokens a day is for the most it can do, not for the whole of what was asked.

Run the Mac mini M4, 24GB at 20M tokens a day · or the Mac mini M6, 24GB

Memory is not what separates them

Every model on this list that fits one machine fits the other, and not only at 32k of context: across all 39 models the calculator counts, at every context from 4k to 256k, there is no model one holds and the other does not. Both leave 16 GB to the GPU. So the choice between them is speed, price and power, not what they can hold.

The assumptions behind both columns

Both columns use the same usage: 500k tokens a day at 15:1 input to output, 32k of context, $0.17 per kWh, and today's API prices held flat. Speeds marked estimated are worked out from memory bandwidth rather than measured, and pay-back scales with them. Graphics cards are priced as the card alone, so add the PC around one before comparing it with a complete computer. Change any of it in the calculator.

More head to head: everything the Mac mini M4, 24GB runs · everything the Mac mini M6, 24GB runs · every other match-up · the quickest pay-back at each level of use · every model against the frontier