Flux is decent at maps - but it has a strong preference for 3d or isometric views. This model will help steer your prompt consistently toward a flat traditional cartography style.

It's reasonably flexible in terms of producing more monochromatic or more colorful maps. It can render decent star maps or otherworldly planetary maps. Trained on traditional, real-world, hand-drawn maps, it brings a believable analog aesthetic.

Now with a new deAnnotated version

Although it will still render text when asked for, and sometimes when you don't, this version is capable of maps with no place names and text. My thanks (and curses, somewhat - it was tedious editing a handful of old maps to blank out all the writing!) to @clydeleegrahamAI for suggesting such. At 1.0 strength, it's very strong and inflexible. All maps end up looking monochrome and a bit simplistic and boring. But at 0.6 to 0.7 you can often get decent maps deAnnotated.

It is hit-and-miss, how much text you get. Obviously Flux understands what maps are already and wants to render some text. So you will likely need to output a number of prompt variations to get what you're looking for.

I still largely prefer the original for its greater versatility, but there's definitely utility in being able to make a map with no (or at least, a lot less) text on it, so I'm happy to have given it a shot and to share it here.