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

What hardware do you need to run MiniMax M2.7?

MiniMax M2.7 at Q4_K_M is 140 GB of weights, with a context ceiling of 200k tokens. Not yet rated: released after our last ratings pass. 10B active out of 229B; fast for its size once it fits.

Cheapest machine that runs itMac Studio M5 Ultra, 256GB at $10,799
Fastest of the ones listedMac Studio M5 Ultra, 256GB — 25 tok/s at 32k context
Honest answer on costPays back in 272 years at 500k tokens a day.

How good is it, really?

On the Artificial Analysis Intelligence Index v4.3 it scores 23, which puts it in the Haiku-class band. In the same band as Anthropic's cheap, fast tier. Every current OpenAI model scores above this band. Score source. See the whole table.

What it costs either way

Renting the same model costs $0.21 per million input tokens and $0.84 per million output (OpenRouter, cheapest active endpoint, checked 2026-09-15). Buying a machine only beats that if you use it hard enough, for long enough, that the hardware price divides down below the rental bill.

Machines that run it

MachinePriceSpeed at 32kPay-back
Mac Studio M5 Ultra, 256GB $10,799 25 tok/s estimated Pays back in 272 years Run the numbers

One machine per family, cheapest first. Speeds are measured where a public benchmark exists and estimated from memory bandwidth otherwise; the calculator says which for any configuration.

The specifics

Parameters
228.7B, of which 10B are active per token
Quantisation
Q4_K_M
Weights on disk
140 GB
KV cache
8.3 GB at 32k context — All 62 layers are full attention in this release, so the context is expensive despite the small active-parameter count.
Maximum context
200k tokens (200k)
Licence
other (see repo)
Sources
source 1