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

What hardware do you need to run Llama 3.1 8B Instruct?

Llama 3.1 8B Instruct at Q8_0 is 8.5 GB of weights, with a context ceiling of 128k tokens. Same model at higher precision. Marginal quality gain; twice the memory.

Cheapest machine that runs itMac mini M6, 24GB at $1,099
Shortest pay-backMac Studio M5 Max, 48GB — Pays back in 1,820 years
Fastest of the ones listedMac Studio M5 Ultra, 96GB — 70 tok/s at 32k context
Honest answer on costPays back in 3,105 years at 500k tokens a day.

How good is it, really?

On the Artificial Analysis Intelligence Index v4.3 it scores 7, which puts it in the Below every hosted tier band. Fine for simple, well-specified tasks. Noticeably less capable than anything the big labs sell today. Score source. See the whole table.

What it costs either way

Renting the same model costs $0.02 per million input tokens and $0.04 per million output (OpenRouter, cheapest active endpoint, checked 2026-09-03). 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 mini M6, 24GB $1,099 9.9 tok/s estimated Pays back in 3,105 years Run the numbers
Strix Halo Framework Desktop, 32GB $1,269 15 tok/s estimated Never pays back Run the numbers
MacBook Pro M5 (14-inch), 32GB $2,399 8.9 tok/s estimated Never pays back Run the numbers
Mac Studio M5 Max, 36GB $2,499 27 tok/s estimated Pays back in 2,568 years Run the numbers
DGX Spark GB10 Grace Blackwell, 128GB $4,699 16 tok/s estimated Never pays back 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
8B
Quantisation
Q8_0
Weights on disk
8.5 GB
KV cache
4.3 GB at 32k context
Maximum context
128k tokens (128k)
Licence
Llama 3.1 Community License
Sources
source 1, source 2