Mac Studio M5 Ultra, 96GB vs Strix Halo HP Z2 Mini G1a, 128GB for local AI
The HP Z2 Mini G1a, 128GB holds 41 of the 50 current models here, and the Mac Studio M5 Ultra, 96GB holds 36. The Mac Studio M5 Ultra, 96GB costs $45 less. On Qwen3.8 27B, the strongest model both hold, the Mac Studio M5 Ultra, 96GB is about 4.8× faster: 48 tok/s against 10, both estimated from memory bandwidth. The Mac Studio M5 Ultra, 96GB pays for itself sooner, in 307 years against 396 years at 500k tokens a day.
| Mac Studio M5 Ultra, 96GB | Strix Halo HP Z2 Mini G1a, 128GB | |
|---|---|---|
| Price | $5,499 | $5,544 |
| Memory | 96 GB | 128 GB |
| Usable by the GPU | 72 GB | 96 GB |
| Memory bandwidth | 1200 GB/s | 256 GB/s |
| Power under load | 270 Wstand-in | 133 Wstand-in |
| Models that fit | 36 | 41 |
| Best model it runs | Qwen3.8 27B | Qwen3.8 27B |
| Speed on that model | 48 tok/s estimated | 10 tok/s estimated |
| Pay-back on that model | Pays back in 307 years | Pays back in 396 years |
Neither power figure above is measured on the machine beside it. Both are stand-ins borrowed from the nearest hardware the data does have, so the gap between them is not a difference between these two machines. Each machine's page names the figure it borrows and why. Every pay-back figure on this page prices its electricity from these numbers.
Run the numbers on the Mac Studio M5 Ultra, 96GB · or the Strix Halo HP Z2 Mini G1a, 128GB
Jump to: The same money, two different machines · How much use it takes to pay back · What the extra memory buys · The assumptions behind both columns
The same money, two different machines
These two cost within $45 of each other: $5,499 for the Mac Studio M5 Ultra, 96GB and $5,544 for the HP Z2 Mini G1a, 128GB. Apple makes one and HP the other. Every other head-to-head on this site holds a piece of the hardware equal and asks what the price gap buys. This one holds the price, so the row that usually carries the answer is the row the two machines agree on, and everything under it is what the same money buys twice.
How much use it takes to pay back
Everything above is at 500k tokens a day. Pay-back moves with how much you actually run, so here are both machines on Qwen3.8 27B, the strongest model both hold, at the five levels of use the calculator names. The Mac Studio M5 Ultra, 96GB pays back sooner at every level of use, so this is not a choice that turns on how hard you work it.
| A day's use | Mac Studio M5 Ultra, 96GB | Strix Halo HP Z2 Mini G1a, 128GB |
|---|---|---|
| 50ka few chats a day | 3,066 years | 3,961 years |
| 200klight assistant use | 766 years | 990 years |
| 1Ma moderate coding-assistant day | 153 years | 198 years |
| 4Mheavy coding with an agent | 38 years | 50 years |
| 20Magents running most of the day | 7.7 years | 14 yearsits ceiling |
On Qwen3.8 27B the Strix Halo HP Z2 Mini G1a, 128GB generates at most 14.3M tokens a day, so its figure at 20M tokens a day is for the most it can do, not for the whole of what was asked.
Run the Mac Studio M5 Ultra, 96GB at 20M tokens a day · or the Strix Halo HP Z2 Mini G1a, 128GB
What the extra memory buys
The Strix Halo HP Z2 Mini G1a, 128GB holds 5 models the Mac Studio M5 Ultra, 96GB cannot at 32k of context. The strongest of them are what the difference in memory actually buys.
| Model | Weights | Needs at 32k | On the Strix Halo HP Z2 Mini G1a, 128GB |
|---|---|---|---|
| Ling 3.0 flashHaiku-class | 78 GB | 78 GB | 41 tok/s estimated |
| Qwen3.5 122B-A10BHaiku-class | 78 GB | 79 GB | 20 tok/s estimated |
| Mistral Small 4 (119B-2603)Below every hosted tier | 74 GB | 75 GB | 32 tok/s estimated |
| Devstral 2 123BBelow every hosted tier | 75 GB | 87 GB | 2.2 tok/s estimated |
| Ornith 1.5 122Bnot yet placed | 76 GB | 77 GB | 21 tok/s estimated |
The extra memory buys context as well. Of the 36 current models both machines hold at 32k, one runs to a longer window on the Strix Halo HP Z2 Mini G1a, 128GB: the weights are a fixed size and the KV cache is not, so what the weights leave spare is what a longer context grows into. Nanbeige4.3 33B reaches 256k there against 128k on the Mac Studio M5 Ultra, 96GB, each the longest window the calculator offers that the machine still holds it at.
The Mac Studio M5 Ultra, 96GB is also head to head with another computer: MacBook Pro M5 Max (16-inch), 64GB. With the same box and the bigger chip: 256GB. With the machine it replaced: Mac Studio M3 Ultra, 96GB.
The HP Z2 Mini G1a, 128GB is also head to head with another computer: Framework Desktop, 128GB.
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. Change any of it in the calculator.
More head to head: everything the Mac Studio M5 Ultra, 96GB runs · everything the Strix Halo HP Z2 Mini G1a, 128GB runs · every other match-up · the quickest pay-back at each level of use · every model against the frontier