Qwen3 235B-A22B Instruct 2507 vs MiniCPM5 2B
Qwen3 235B-A22B Instruct 2507 and MiniCPM5 2B both score 13 on the intelligence index. The cheapest machine here that runs Qwen3 235B-A22B Instruct 2507 is the Mac Studio M5 Ultra, 256GB, at $10,799. MiniCPM5 2B runs on the Mac mini M6, 16GB at $899, $9,900 less. On the Mac Studio M5 Ultra, 256GB, the cheapest machine here that runs both, MiniCPM5 2B is about 16.8× quicker: 303 tok/s against 18, both estimated from memory bandwidth. At 500k tokens a day the Mac Studio M5 Ultra, 256GB pays for itself in 939 years running MiniCPM5 2B, against 977 years running Qwen3 235B-A22B Instruct 2507.
| Qwen3 235B-A22B Instruct 2507 | MiniCPM5 2B | |
|---|---|---|
| Intelligence index | 13 | 13 |
| Class | Below every hosted tier | Below every hosted tier |
| Weights | 142 GB | 1.6 GB |
| Needs at 32k | 148 GB | 3.0 GB |
| Quantisation | Q4_K_M | Q4_K_M |
| Parameters | 235.1B (22B active) | 2.5B |
| Max context | 256k | 128k |
| API price per 1M | $0.0875 in / $0.35 out | $0.06 in / $0.15 outpriced as Granite 4.2 8B |
| Licence | Apache 2.0 | Apache 2.0 |
| Machines here that run it | 1 of 37 | 37 of 37 |
| Cheapest machine that runs it | Mac Studio M5 Ultra, 256GB $10,799 | Mac mini M6, 16GB $899 |
Run Qwen3 235B-A22B Instruct 2507 on the Mac Studio M5 Ultra, 256GB · or MiniCPM5 2B on the Mac mini M6, 16GB
Jump to: Side by side on the Mac Studio M5 Ultra, 256GB · How much use it takes to pay for the machine · Machines that run one and not the other · The assumptions behind both columns
Side by side on the Mac Studio M5 Ultra, 256GB
The table above gives each model the cheapest machine that runs it, and those are two different machines, so nothing in it is a like-for-like race. The Mac Studio M5 Ultra, 256GB is the cheapest machine here that runs both, so this is the pair doing the same work on the same hardware.
| Qwen3 235B-A22B Instruct 2507 | MiniCPM5 2B | |
|---|---|---|
| Speed at 32k | 18 tok/s estimated | 303 tok/s estimated |
| Pay-back on this machine | Pays back in 977 years | Pays back in 939 years |
| API cost per month | $1.58 | $1.00priced as Granite 4.2 8B |
Run Qwen3 235B-A22B Instruct 2507 on the Mac Studio M5 Ultra, 256GB · or MiniCPM5 2B
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. MiniCPM5 2B 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 use | Qwen3 235B-A22B Instruct 2507 | MiniCPM5 2B |
|---|---|---|
| 50ka few chats a day | 9,774 years | 9,387 years |
| 200klight assistant use | 2,444 years | 2,347 years |
| 1Ma moderate coding-assistant day | 489 years | 469 years |
| 4Mheavy coding with an agent | 122 years | 117 years |
| 20Magents running most of the day | 24 years | 23 years |
Run Qwen3 235B-A22B Instruct 2507 at 20M tokens a day · or MiniCPM5 2B
Machines that run one and not the other
MiniCPM5 2B needs 3.0 GB of memory at 32k of context and Qwen3 235B-A22B Instruct 2507 needs 148 GB. That puts MiniCPM5 2B on 36 of the 37 machines priced here that Qwen3 235B-A22B Instruct 2507 does not, starting at $899.
| Machine | Price | Memory | Speed on MiniCPM5 2B | Pay-back |
|---|---|---|---|---|
| Mac mini M6, 16GB | $899 | 16 GB | 39 tok/s estimated | Pays back in 81 years |
| Mac mini M6, 24GB | $1,099 | 24 GB | 43 tok/s estimated | Pays back in 98 years |
| Strix Halo Framework Desktop, 32GB | $1,269 | 32 GB | 65 tok/s estimated | Pays back in 117 years |
| Mac mini M6, 32GB | $1,299 | 32 GB | 43 tok/s estimated | Pays back in 116 years |
| MacBook Air M5 (13-inch), 16GB | $1,299 | 16 GB | 39 tok/s estimated | Pays back in 117 years |
| AMD Radeon AI PRO R9700, 32GB | $1,299card only | 32 GB | 188 tok/s estimated | Pays back in 117 years |
30 more, on the MiniCPM5 2B page.
Qwen3 235B-A22B Instruct 2507 is also head to head with Nemotron 3.5 Lightning 30B-A3B above it on the leaderboard and Qwen3.8 Flash Next, the current Qwen nearest it in size. One more model needs much the same memory: MiniMax M2.7. MiniCPM5 2B is also head to head with Qwen3.5 4B below it on the leaderboard.
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 Qwen3 235B-A22B Instruct 2507 · every machine that runs MiniCPM5 2B · every other match-up · both against the frontier · the quickest pay-back at each level of use