If you've been following the world of AI-generated imagery, the last few weeks have been hard to ignore. Midjourney just wrapped up the final round of its image rating party on Discord and at midjourney.com/rank-v8 — and that's always one signal: a model launch is right around the corner. V8 is shaping up to be the biggest update in the platform's history. Not an evolution. A revolution.



A Brief History of Midjourney Models – Where We've Come From


Midjourney launched in July 2022 as a niche research project based in San Francisco, founded by David Holz. The earliest versions of the model were fascinating but rough — they generated dreamlike, often blurry images with limited control over detail.


V1–V3 (2022) — Experimental, dreamy aesthetics, very limited coherence.

V4 (2022–2023) — A breakthrough moment. The model began understanding complex prompts, with improved compositional consistency and better human anatomy.

V5 / V5.2 (2023) — Midjourney started seriously competing for the title of best AI image generator. V5.2 introduced the Vary (Region) feature — local editing of selected areas without affecting the rest of the image.

V6 (December 2023) — Significantly better interpretation of long, detailed prompts, improved coherence and world knowledge.

V6.1 (July 2024) — The default model until June 2025. More precise details and textures, generation roughly 25% faster than V6. The web UI launched alongside it.

V7 (April 2025, default from June 2025) — A leap forward in rendering hands and bodies, Draft Mode, Omni Reference. But the team itself acknowledged that V7 didn't fully meet expectations.

V8 (launching any day now) — A full rewrite. New code, new architectures, new datasets.


What We Know About Midjourney V8 – The Key New Features


🖼️ Native 2K Resolution

This is probably the biggest visual change. Previous versions generated images at lower resolutions, and higher quality required post-generation upscaling — which could introduce artifacts or soften details. V8 natively supports resolutions from 64 all the way up to 2048 pixels (2K) and beyond — no upscaler in the loop. Sharper. Cleaner. Right out of the box.


✍️ Finally: Readable Text in Images

Generating legible text inside AI images has been an industry-wide pain point for years. Road signs, product labels, poster typography — everything came out with typos or a blurry mess of characters. V8 directly targets this problem, delivering clear and contextually accurate text rendering. Round 2 of the rating party was dedicated entirely to testing typography — Midjourney collected ratings on images containing text specifically to tune the model before launch. This is huge news for designers creating ads, branding materials, and mockups.


🏗️ Complete Architectural Overhaul

V8 isn't a patch — it's a full rewrite of the training infrastructure. New code, new architectures, new datasets. What does this mean in practice? Faster feature development after launch. New capabilities will be able to ship much more quickly than before. The Midjourney team has been explicit that one of the core motivations for the full rewrite was enabling faster iteration going forward.


🎯 Better Prompt Understanding

Advanced natural language understanding will allow for better interpretation of complex, multi-element prompts. Fewer iterations, more first-try wins.


🔧 Improved Local Editing Tools

The architectural rewrite also brings better local editing tools — more precise face rendering and the ability to make targeted corrections to specific areas of an image without affecting the overall composition.


⚡ Performance Optimization

Faster generation times and reduced operational costs compared to previous versions — though high demand at launch may result in temporary tier-based access limits.



The Rating Party – How Midjourney Builds Models With Its Community


This is one of the most unique aspects of Midjourney: the company actively involves its community in the model training process.


Before V8 reaches everyone, Midjourney ran a multi-round image rating process. At midjourney.com/rank-v8, users are shown two images at a time and click the one they find more aesthetically pleasing — or skip if they're unsure. On a computer, you can use the 1 and 2 keys.


The rating system serves two key purposes. First, learning what is "bad" — early rounds deliberately showed plain, boring, or even outright bad images. The data from these sessions teaches the model what users don't want, which is just as important as learning what they do. Second, personalization calibration — the final round shows images with strongly contrasting aesthetics. Based on this, V8 calibrates its personalization system to better match generation to each individual user's taste.


Three rounds of the rating party took place between February 14th and 20th, 2026. The Final Round is running all the way through to launch.


Android patrzy z góry człowiekowi w oczy

Czy modele są świadome?



When Does Midjourney V8 Launch?


Based on official announcements and David Holz's comments during Office Hours on February 4th and 11th, 2026 — the launch is planned for late February or early March 2026. The team is in a "quiet polishing phase," training at 1024×1024 resolution is complete, and internal and moderator testing is already underway.

This isn't another vague "coming soon." It's genuinely any day now.



How to Prepare for V8


Join the rating party — midjourney.com/rank-v8 is still active. Your ratings are literally shaping the final version of the model.


Save your best SREF codes — style codes should transfer well to V8, so keep them handy.


Build prompts around text — if you've been avoiding text in images because the results were unreliable, V8 changes that. Time to prepare prompts for ads, branding, and typography work.


Follow official channels — updates are posted at updates.midjourney.com and on the Midjourney Discord server.



Summary


Midjourney V8 is not just another version with a handful of tweaks. It's a fundamental rewrite of a platform that has defined the standards of AI-generated imagery for the past three years. Native 2K resolution, readable text in images, a new architecture, and deeply integrated personalization — this is a set of changes that could genuinely transform how designers, content creators, and digital artists work.

The final rating party is live. The launch is closing in fast. If you don't have a Midjourney account yet — now's a good time to change that.


Post updated: March 2026. Sources: official Midjourney announcements (updates.midjourney.com), Midjourney Office Hours February 2026, Geeky Gadgets, RealAIGirls.