Five models with different needs
The model choices range from a 4.7 GB FLUX.2 Klein download to the 22.2 GB Krea 2 Q8 option. Model Manager shows storage and memory guidance so you can choose what fits your Mac.
Measured on M5 Max
In a 1024 px test on a MacBook Pro with M5 Max and 48 GB unified memory, FLUX.2 Klein took 8.4 seconds. The other models took between 22.8 and 32 seconds, depending on model and Fast mode.
Your Mac does the work
The model uses the Mac to process the prompt and make the image. That keeps the image workflow local and removes the need for cloud generation credits or a separate account.
Apple silicon Macs use one shared pool of memory for normal tasks and graphics work. Local image generation uses part of that pool, so closing a large game, video editor, or browser session can help when the Mac is low on available memory. The app gives memory guidance for each model before you download it.
Generation speed depends on more than the name of the chip. Image size, model choice, settings, adapters, available memory, and other open apps can all change the time. A benchmark is useful for rough comparison, but it is not a promise that every Mac or every prompt will produce the same number.
Start with a smaller model if you are unsure what the Mac can handle. Make a square image at a normal size, then increase the size or add adapters after the basic workflow is stable. This makes it easier to tell whether a slowdown comes from the model, the chosen settings, or other work happening on the computer.
Radiant Canvas