Professional
The one that runs the models people actually want.
128 GB of unified memory, which is the line between running a capable assistant and running the frontier of what is openly available. For developers, researchers, and teams who have stopped experimenting.
£2,899 excluding VAT, installation included · UK orders add 20% VAT at checkout
£250 deposit · fully refundable · nothing charged until it ships
What the extra memory buys
Memory is not speed. It is permission.
A bigger machine does not make a small model faster. It decides which models will load at all, and that is a hard wall rather than a slowdown.
Models Personal cannot load
A 70B model needs about 40 GB plus context headroom. It will not open on 64 GB. On 128 GB it runs with room to spare, and the quality difference over a 32B is not subtle.
Several models at once
A fast one for everyday questions, a strong one for analysis, a coding model resident alongside them. No swapping, no reload wait.
Long documents
Context is held in memory alongside the model. At 128k tokens that can cost as much again as the model itself, which is exactly the case Personal runs out of room for.
A small team
Every simultaneous conversation needs its own memory. This is the machine that serves five people without them queueing behind each other.
Specification
What you get
The same operating system, isolation switch and installation as every machine in the range. Twice the memory of Personal, and the ability to link a second unit when that is no longer enough.
- Memory
- 128 GB unified
- Largest model
- 70B comfortably, 120B mixture-of-experts
- Models
- pre-installed on the drive
- Network
- physical isolation switch
- Stacking
- links to a second unit for 256 GB pooled
- Installation
- included, by a SetupMyAI engineer
Connections
Everything plugs in at the back
Four Thunderbolt ports, two 10 gigabit ethernet, USB-A, HDMI and DisplayPort. Four rather than two so external storage, displays and capture can all be connected at once instead of taking turns. Nothing here is needed for stacking: that happens through the feet. The vent band runs the full width because the machine is cooled by moving air, not by throttling.
Straight answer
Buy Personal instead if this is you
- You are one person, and a 32B model answers your questions well enough. It genuinely will for most work.
- You are testing whether local AI is useful at all. Learn on the cheaper machine first.
- You do not work with long documents or large codebases routinely.
Personal is £1,500 and runs the same software. If you outgrow it you can link a second machine rather than replacing it, so buying the smaller one first is not a wasted decision.
Not sure which one?
Answer three questions and we will tell you, including when the cheaper machine is the right one.