kayetana
Comparison

In your office or in a data center?

A rented GPU server brings lots of compute, but it sits somewhere else. A Mac Studio sits with you, quiet and smaller than a shoebox.

kayetana.local
Installed models with size, context and capabilities
At a glance

Owning hardware versus renting.

Mac Studio in your officeRented GPU server
Where data livesYour office, your networkThe host's data center
Works offline✓–
AccessOn your own networkOver the internet
Very large modelsUp to the Mac Studio's memoryBigger GPUs available to rent
CostHardware once, flat platform feeOngoing rent
Hardware failureYour job, nightly backupsHandled by the host
Space and noiseQuiet, fits on a deskNo office space needed
SetupScript for models, HTTPS, backupsDepends on the offer

Rental offers vary a lot. This column describes the typical model, not any specific provider.

Hardware

One Mac Studio is enough for your team.

We recommend a Mac Studio with 36 GB of memory or more. That runs models up to about 20 billion parameters alongside search and transcription. More memory fits larger models.

  • qwen3:4b-instruct, qwen3:14b and gpt-oss:20b for chat and reasoning
  • bge-m3 for search, rerank and classification
  • Whisper large-v3-turbo for speech
Weighing it up

What speaks for each side.

Mac Studio: location

The hardware sits in your office. Data never leaves the building.

Mac Studio: offline

Once the models are loaded, it runs without the internet.

Mac Studio: no rent

You buy the hardware once instead of paying rent every month.

Renting: large models

Big GPUs run models that need more memory than a Mac Studio has.

Renting: no hardware

The host takes care of failures and replacements.

Renting: from anywhere

A server in a data center can be reached over the internet.

Which hardware fits you?

We'll help you pick the right Mac Studio and show you kayetana in a demo.

FAQ

Frequently asked questions

Which Mac Studio do I need?

We recommend 36 GB of memory or more. That runs models up to about 20 billion parameters alongside search and transcription. More memory allows larger models.

Does kayetana run on a rented server too?

kayetana is built for a Mac Studio with Apple silicon on your own network. Talk to us if you have a different environment in mind.

What happens if the Mac Studio fails?

kayetana backs up the database and files every night at 3:15 a.m. and keeps the backups for 14 days. We also recommend an external backup.

Do I have to take care of updates?

You install updates with a script. After a reboot, all services start on their own.