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

What hardware do you need to run Qwen3 235B-A22B Instruct 2507?

Qwen3 235B-A22B Instruct 2507 at Q4_K_M is 142 GB of weights, with a context ceiling of 256k tokens. The closest thing to a frontier model that fits in 256 GB. Needs an Ultra-class Mac.

Cheapest machine that runs itMac Studio M5 Ultra, 256GB at $10,799
Fastest of the ones listedMac Studio M5 Ultra, 256GB — 18 tok/s at 32k context
Honest answer on costPays back in 977 years at 500k tokens a day.

How good is it, really?

On the Artificial Analysis Intelligence Index v4.3 it scores 13 (reasoning mode; 12 without), 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 $0.0875 per million input tokens and $0.35 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 Studio M5 Ultra, 256GB $10,799 18 tok/s estimated Pays back in 977 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
235.1B, of which 22B are active per token
Quantisation
Q4_K_M
Weights on disk
142 GB
KV cache
6.3 GB at 32k context
Maximum context
256k tokens (256k natively)
Licence
Apache 2.0
Sources
source 1, source 2