Granite 4.2 8B vs Qwen3 8B
Granite 4.2 8B scores higher on the intelligence index, 12 against 5. The cheapest machine here that runs Granite 4.2 8B is the Mac mini M6, 24GB, at $1,099. Qwen3 8B runs on the Mac mini M6, 16GB at $899, $200 less. On the Mac mini M6, 24GB, the cheapest machine here that runs both, Qwen3 8B is about 1.1× quicker: 13 tok/s against 12, both estimated from memory bandwidth. At 500k tokens a day the Mac mini M6, 24GB pays for itself in 45 years running Qwen3 8B, against 122 years running Granite 4.2 8B.
| Granite 4.2 8B | Qwen3 8B | |
|---|---|---|
| Intelligence index | 12 | 5 |
| Class | Below every hosted tier | Below every hosted tier |
| Weights | 5.3 GB | 5.0 GB |
| Needs at 32k | 11 GB | 9.9 GB |
| Quantisation | Q4_K_M | Q4_K_M |
| Parameters | 8B | 8.2B |
| Max context | 128k | 40k |
| API price per 1M | $0.06 in / $0.15 out | $0.15 in / $0.15 outpriced as Ministral 3 8B |
| Licence | Apache 2.0 | Apache 2.0 |
| Machines here that run it | 33 of 37 | 37 of 37 |
| Cheapest machine that runs it | Mac mini M6, 24GB $1,099 | Mac mini M6, 16GB $899 |
Run Granite 4.2 8B on the Mac mini M6, 24GB · or Qwen3 8B on the Mac mini M6, 16GB
Jump to: Side by side on the Mac mini M6, 24GB · 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 mini M6, 24GB
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 mini M6, 24GB is the cheapest machine here that runs both, so this is the pair doing the same work on the same hardware.
| Granite 4.2 8B | Qwen3 8B | |
|---|---|---|
| Speed at 32k | 12 tok/s estimated | 13 tok/s estimated |
| Pay-back on this machine | Pays back in 122 years | Pays back in 45 years |
| API cost per month | $1.00 | $2.28priced as Ministral 3 8B |
Run Granite 4.2 8B on the Mac mini M6, 24GB · or Qwen3 8B
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, 24GB. Qwen3 8B 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 | Granite 4.2 8B | Qwen3 8B |
|---|---|---|
| 50ka few chats a day | 1,216 years | 445 years |
| 200klight assistant use | 304 years | 111 years |
| 1Ma moderate coding-assistant day | 61 years | 22 years |
| 4Mheavy coding with an agent | 15 years | 5.6 years |
| 20Magents running most of the day | 3.7 yearsits ceiling | 15 monthsits ceiling |
The Mac mini M6, 24GB cannot generate 20M tokens a day on either model: at most 16.4M on Granite 4.2 8B and 17.9M on Qwen3 8B. Both figures on that row are for the most it can do.
Run Granite 4.2 8B at 20M tokens a day · or Qwen3 8B
Machines that run one and not the other
Qwen3 8B needs 9.9 GB of memory at 32k of context and Granite 4.2 8B needs 11 GB. That puts Qwen3 8B on 4 of the 37 machines priced here that Granite 4.2 8B does not, starting at $899.
| Machine | Price | Memory | Speed on Qwen3 8B | Pay-back |
|---|---|---|---|---|
| Mac mini M6, 16GB | $899 | 16 GB | 12 tok/s estimated | Pays back in 37 years |
| MacBook Air M5 (13-inch), 16GB | $1,299 | 16 GB | 12 tok/s estimated | Pays back in 53 years |
| MacBook Air M5 (15-inch), 16GB | $1,499 | 16 GB | 12 tok/s estimated | Pays back in 61 years |
| MacBook Pro M5 (14-inch), 16GB | $1,999 | 16 GB | 12 tok/s estimated | Pays back in 82 years |
Granite 4.2 8B is also head to head with Ling 3.0 tiny above it on the leaderboard and GLM-4.5-Air below it. Qwen3 8B is also head to head with Ministral 3 14B above it on the leaderboard, Gemma 3 27B it below it and Qwen3.5 9B, the current Qwen nearest it in size. One more model needs much the same memory: Gemma 3 12B it.
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 Granite 4.2 8B · every machine that runs Qwen3 8B · every other match-up · both against the frontier · the quickest pay-back at each level of use