Gemma 4 12B vs Mellum2.1 12B-A2.5B Thinking
Gemma 4 12B scores 14 on the intelligence index, and Mellum2.1 12B-A2.5B Thinking has not been placed on it. Both take the same machine to start: the cheapest here that runs either is the Mac mini M6, 16GB, at $899. On it, Mellum2.1 12B-A2.5B Thinking is about 1.5× quicker: 21 tok/s against 14, both estimated from memory bandwidth. At 500k tokens a day the Mac mini M6, 16GB pays for itself in 67 years running Mellum2.1 12B-A2.5B Thinking, against 71 years running Gemma 4 12B.
| Gemma 4 12B | Mellum2.1 12B-A2.5B Thinking | |
|---|---|---|
| Intelligence index | 14 | not placed |
| Class | Below every hosted tier | not yet placed |
| Weights | 7.1 GB | 8.1 GB |
| Needs at 32k | 8.0 GB | 8.6 GB |
| Quantisation | Q4_K_M | Q4_K_M |
| Parameters | 12B | 12.2B (2.5B active) |
| Max context | 256k | 128k |
| API price per 1M | $0.08 in / $0.12 outpriced as Qwen3.5 9B | $0.08 in / $0.12 outpriced as Qwen3.5 9B |
| Licence | Apache 2.0 | Apache 2.0 |
| Machines here that run it | 37 of 37 | 37 of 37 |
| Cheapest machine that runs it | Mac mini M6, 16GB $899 | Mac mini M6, 16GB $899 |
Run Gemma 4 12B on the Mac mini M6, 16GB · or Mellum2.1 12B-A2.5B Thinking
Jump to: Side by side on the Mac mini M6, 16GB · How much use it takes to pay for the machine · Memory is not what separates them · The assumptions behind both columns
Side by side on the Mac mini M6, 16GB
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 mini M6, 16GB, at $899.
| Gemma 4 12B | Mellum2.1 12B-A2.5B Thinking | |
|---|---|---|
| Speed at 32k | 14 tok/s estimated | 21 tok/s estimated |
| Pay-back on this machine | Pays back in 71 years | Pays back in 67 years |
| API cost per month | $1.26priced as Qwen3.5 9B | $1.26priced as Qwen3.5 9B |
Run Gemma 4 12B on the Mac mini M6, 16GB · or Mellum2.1 12B-A2.5B Thinking
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 mini M6, 16GB. Mellum2.1 12B-A2.5B Thinking 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 | Gemma 4 12B | Mellum2.1 12B-A2.5B Thinking |
|---|---|---|
| 50ka few chats a day | 712 years | 671 years |
| 200klight assistant use | 178 years | 168 years |
| 1Ma moderate coding-assistant day | 36 years | 34 years |
| 4Mheavy coding with an agent | 8.9 years | 8.4 years |
| 20Magents running most of the day | 22 monthsits ceiling | 20 months |
The Mac mini M6, 16GB generates at most 19.8M tokens a day on Gemma 4 12B, so that column's figure at 20M tokens a day is for the most it can do, not for the whole of what was asked.
Run Gemma 4 12B at 20M tokens a day · or Mellum2.1 12B-A2.5B Thinking
Memory is not what separates them
Gemma 4 12B needs 8.0 GB of memory at 32k of context and Mellum2.1 12B-A2.5B Thinking needs 8.6 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 and how fast it runs, not what you have to buy to hold it.
Gemma 4 12B is also head to head with GLM-4.7-Flash above it on the leaderboard, Qwen3.5 9B below it and Gemma 3 12B it, the last-generation Gemma 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 Gemma 4 12B · every machine that runs Mellum2.1 12B-A2.5B Thinking · every other match-up · both against the frontier · the quickest pay-back at each level of use