Gemma 4 26B-A4B vs Qwen3.5 122B-A10B
Gemma 4 26B-A4B scores higher on the intelligence index, 17 against 16. The cheapest machine here that runs Qwen3.5 122B-A10B is the Framework Desktop, 128GB, at $3,449. Gemma 4 26B-A4B runs on the Framework Desktop, 32GB at $1,269, $2,180 less. On the Framework Desktop, 128GB, the cheapest machine here that runs both, Gemma 4 26B-A4B is about 2.1× quicker: 41 tok/s against 20, both estimated from memory bandwidth. At 500k tokens a day the Framework Desktop, 128GB pays for itself in 53 years running Qwen3.5 122B-A10B, against 434 years running Gemma 4 26B-A4B.
| Gemma 4 26B-A4B | Qwen3.5 122B-A10B | |
|---|---|---|
| Intelligence index | 17 | 16 |
| Class | Haiku-class | Haiku-class |
| Weights | 17 GB | 78 GB |
| Needs at 32k | 18 GB | 79 GB |
| Quantisation | Q4_K_M | UD-Q4_K_M |
| Parameters | 25.2B (3.8B active) | 125.1B (10B active) |
| Max context | 256k | 256k |
| API price per 1M | $0.042 in / $0.22 out | $0.26 in / $2.08 out |
| Licence | Apache 2.0 | Apache 2.0 |
| Machines here that run it | 30 of 37 | 12 of 37 |
| Cheapest machine that runs it | Strix Halo Framework Desktop, 32GB $1,269 | Strix Halo Framework Desktop, 128GB $3,449 |
Run Gemma 4 26B-A4B on the Framework Desktop, 32GB · or Qwen3.5 122B-A10B on the Framework Desktop, 128GB
Jump to: Side by side on the Framework Desktop, 128GB · 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 Framework Desktop, 128GB
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 Strix Halo Framework Desktop, 128GB is the cheapest machine here that runs both, so this is the pair doing the same work on the same hardware.
| Gemma 4 26B-A4B | Qwen3.5 122B-A10B | |
|---|---|---|
| Speed at 32k | 41 tok/s estimated | 20 tok/s estimated |
| Pay-back on this machine | Pays back in 434 years | Pays back in 53 years |
| API cost per month | $0.81 | $5.69 |
Run Gemma 4 26B-A4B on the Framework Desktop, 128GB · or Qwen3.5 122B-A10B
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 Framework Desktop, 128GB. Qwen3.5 122B-A10B 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 26B-A4B | Qwen3.5 122B-A10B |
|---|---|---|
| 50ka few chats a day | 4,335 years | 533 years |
| 200klight assistant use | 1,084 years | 133 years |
| 1Ma moderate coding-assistant day | 217 years | 27 years |
| 4Mheavy coding with an agent | 54 years | 6.7 years |
| 20Magents running most of the day | 11 years | 16 months |
Run Gemma 4 26B-A4B at 20M tokens a day · or Qwen3.5 122B-A10B
Machines that run one and not the other
Gemma 4 26B-A4B needs 18 GB of memory at 32k of context and Qwen3.5 122B-A10B needs 79 GB. That puts Gemma 4 26B-A4B on 18 of the 37 machines priced here that Qwen3.5 122B-A10B does not, starting at $1,269.
| Machine | Price | Memory | Speed on Gemma 4 26B-A4B | Pay-back |
|---|---|---|---|---|
| Strix Halo Framework Desktop, 32GB | $1,269 | 32 GB | 41 tok/s estimated | Pays back in 160 years |
| Mac mini M6, 32GB | $1,299 | 32 GB | 15 tok/s estimated | Pays back in 177 years |
| AMD Radeon AI PRO R9700, 32GB | $1,299card only | 32 GB | 97 tok/s estimated | Pays back in 162 years |
| Strix Halo Corsair AI Workstation 300, 64GB | $1,700 | 64 GB | 41 tok/s estimated | Pays back in 214 years |
| Strix Halo Framework Desktop, 64GB | $1,959 | 64 GB | 41 tok/s estimated | Pays back in 246 years |
| NVIDIA GeForce RTX 5090, 32GB | $1,999card only | 32 GB | 172 tok/s estimated | Pays back in 253 years |
12 more, on the Gemma 4 26B-A4B page.
Gemma 4 26B-A4B is also head to head with Muse Glimmer 30B above it on the leaderboard. Qwen3.5 122B-A10B is also head to head with Gemma 4 31B it below it on the leaderboard. Two more models need much the same memory: Ling 3.0 flash and GLM-4.5-Air.
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 26B-A4B · every machine that runs Qwen3.5 122B-A10B · every other match-up · both against the frontier · the quickest pay-back at each level of use