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

MiniMax M2.7 vs Qwen3 235B-A22B Instruct 2507

MiniMax M2.7 scores higher on the intelligence index, 23 against 13. Both take the same machine to start: the cheapest here that runs either is the Mac Studio M5 Ultra, 256GB, at $10,799. On it, MiniMax M2.7 is about 1.4× quicker: 25 tok/s against 18, both estimated from memory bandwidth. At 500k tokens a day the Mac Studio M5 Ultra, 256GB pays for itself in 272 years running MiniMax M2.7, against 977 years running Qwen3 235B-A22B Instruct 2507.

MiniMax M2.7Qwen3 235B-A22B Instruct 2507
Intelligence index2313
ClassHaiku-classBelow every hosted tier
Weights140 GB142 GB
Needs at 32k149 GB148 GB
QuantisationQ4_K_MQ4_K_M
Parameters228.7B (10B active)235.1B (22B active)
Max context200k256k
API price per 1M$0.21 in / $0.84 out$0.0875 in / $0.35 out
Licenceother (see repo)Apache 2.0
Machines here that run it1 of 371 of 37
Cheapest machine that runs itMac Studio M5 Ultra, 256GB $10,799Mac Studio M5 Ultra, 256GB $10,799
Summarising not rated good
Translation not rated good
Everyday coding not rated good
Reasoning & maths not rated good
Agentic work not rated usable

Run MiniMax M2.7 on the Mac Studio M5 Ultra, 256GB · or Qwen3 235B-A22B Instruct 2507

Ratings are coarse on purpose: they say what a model is usable for, not where it places to the decimal.

Side by side on the Mac Studio M5 Ultra, 256GB

The cheapest machine that runs either model is the same one, so this is the pair doing the same work on the same hardware: Mac Studio M5 Ultra, 256GB, at $10,799.

MiniMax M2.7Qwen3 235B-A22B Instruct 2507
Speed at 32k25 tok/s estimated18 tok/s estimated
Pay-back on this machinePays back in 272 yearsPays back in 977 years
API cost per month$3.80$1.58

Run MiniMax M2.7 on the Mac Studio M5 Ultra, 256GB · or Qwen3 235B-A22B Instruct 2507

How much use it takes to pay for the machine

Everything above is at 500k tokens a day. Pay-back moves with how much you actually run, so here are both models at the five levels of use the calculator names, on the Mac Studio M5 Ultra, 256GB. MiniMax M2.7 pays for it sooner at every level of use, so which of them to run does not turn on how hard you work it.

A day's useMiniMax M2.7Qwen3 235B-A22B Instruct 2507
50ka few chats a day2,720 years9,774 years
200klight assistant use680 years2,444 years
1Ma moderate coding-assistant day136 years489 years
4Mheavy coding with an agent34 years122 years
20Magents running most of the day6.8 years24 years

Run MiniMax M2.7 at 20M tokens a day · or Qwen3 235B-A22B Instruct 2507

Memory is not what separates them

MiniMax M2.7 needs 149 GB of memory at 32k of context and Qwen3 235B-A22B Instruct 2507 needs 148 GB. Every machine priced here that runs one runs the other, at 32k of context. So the choice between them is what each is good at, how fast it runs and what the same work costs on an API, not what you have to buy to hold it.

MiniMax M2.7 is also head to head with Ling 3.0 flash above it on the leaderboard and Qwen3.6 27B below it. One more model needs much the same memory: DeepSeek V4-Flash. Qwen3 235B-A22B Instruct 2507 is also head to head with Nemotron 3.5 Lightning 30B-A3B above it on the leaderboard, MiniCPM5 2B below it and Qwen3.8 Flash Next, the current Qwen nearest it in size.

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. Where nobody rents an open model by the token, its API prices are the nearest hosted model's, named beside them. Machines are the 37 here with a published price that are still sold. Change any of it in the calculator.

More head to head: every machine that runs MiniMax M2.7 · every machine that runs Qwen3 235B-A22B Instruct 2507 · every other match-up · both against the frontier · the quickest pay-back at each level of use