Midjourney launches V8 alpha with 5x speed boost and native 2K

If you have spent the last few months twiddling your thumbs while waiting for your Midjourney grids to roll, the wait is about to get a lot shorter.

Midjourney just rolled out the alpha version of its V8 model on alpha.midjourney.com. The headline jump is sheer performance: the team claims base image generation is now roughly 5x faster than V7, paired with tighter prompt obedience and noticeably crisper in-image text when placed inside quotes.

Raw 2K, new web tools, and a heavy price tag

To keep pace with the 5x speed bump, Midjourney overhauled its web interface. There is a new conversation mode for rapid-fire iterative prompting, a dedicated Grid Mode to inspect batches in detail, and tucked-away sidebar settings that stop clogging your view.

On the visual side, V8 introduces an `--hd` flag that natively renders images at 2K resolution, alongside a `--q 4` switch for artists who need extra structural coherence. Old pros will be glad to hear that V7 personalization profiles, moodboards, and sref collections carry over seamlessly, along with staple parameters like `--chaos`, `--weird`, `--exp`, and `--raw`.

There is, however, a catch that will sting your subscription balance. Relax mode is completely disabled for V8 right now while the company builds out dedicated server clusters. To make matters tighter, using `--hd`, `--q 4`, sref, or moodboards slows the generation down by 4x and burns fast hours at a 4x rate.

Midjourney also noted that V8 is still finding its visual footing. The team recommends leaning heavily into personalization—even dialing it all the way up to `--stylize 1000`—or flipping to `--raw` immediately if you want photorealistic and grounded compositions.

Why it matters

We think Midjourney made the right call by prioritizing turnaround speed over gimmick features. Shaving generation time down to a fraction of V7 transforms prompt crafting from an agonizing wait into an active, conversational design loop.

For studio concept artists and commercial illustrators, native 2K rendering removes one more tedious upscaling step from the delivery pipeline. If you already pay for high-tier Fast GPU quotas and need tight typographic signs or hyper-specific scene layouts, jumping into the alpha right now is a no-brainer.

That said, casual creators should probably hold their horses. Burning 4x compute on every styled or HD run without the safety net of Relax mode makes experimentation expensive. Midjourney is essentially asking paying power users to foot the compute bill for model alignment via lightbox rating parties, which is clever for rapid data collection, but might alienate credit-conscious hobbyists until those cheaper clusters come online.