Midjourney has started a test of “Thinking Mode” for image generation, and only people on its alpha site can try it. The company says it “might” offer the feature to everyone later.
Access runs through a finished job. Open the lightbox, then press “Rerun (Thinking)” at the lower right. That makes it a second pass on an image you already have, not a new way to write a prompt.
In a post on its updates page, Caleb of Midjourney says the team is “finding in tests” that the mode makes the model better at following prompts, rendering lettering, and keeping an image coherent. The post offers no measurements and no side-by-side samples. It also never says what “thinking” means inside an image model, so assuming it behaves like a text reasoning model is a guess.
Midjourney wants users to report where the mode helps and where it does not. It floats two possible shapes: a standalone mode, or a way to “add more thinking” after a job has run.
Anyone who bills for text-heavy Midjourney work can rerun a few jobs now, since no price or tier has been mentioned.
Midjourney, updates page post by Caleb, 8 October 2026.
