AI tools·Image & video·Global
Midjourney
A taste benchmark for image gen. Started on Discord, then Web; default around V8.2 in 2026.
Compiled through 2026-09
- Vendor
- Midjourney, Inc.
- Started
- 2022
- China availability
- Overseas accessRequires overseas network access
- Pricing tier
- Paid
- Current flagship
- Midjourney v6.1 / V7
- Website
- www.midjourney.com
Evaluation, Pros & Cons
Editorial score based on public benchmarks and community feedback — not a third-party benchmark result
Key Advantages
- World-class artistic text-to-image quality; top-tier lighting, aesthetics, and detail
- Web creation UI plus advanced inpainting, panning, and aspect-ratio controls
Limitations & Caveats
- Paid subscription only; strict commercial usage tiers and no perpetual free tier
Overview
Midjourney is known for taste, composition, and community aesthetics—concept art, posters, characters. It grew from Discord into the web app. V6/V7 improved realism and text; V8 became the 2026 default (V8.2 reported around July). Video is not the main arena versus Runway / Kling / Sora; still images and controllable style are.
Direction
The cadence is a new default model plus better prompt understanding and consistency. Web editing, character/style reference, and inpainting continue. Video is explored; still-image quality remains the center of gravity.
Version history
Jul 2026
LatestV8.2 default
Reported current default.
View details
New capabilities
- Reported default cutover to V8.2
What got stronger
- Another tick on detail, consistency, and prompts
Version notes
V8.2 shows up in mid-2026 public discussion as the default. Video is still not home turf. Trust the official model list.
Apr 2025
V7
Next default model.
View details
New capabilities
- V7 as the next default
What got stronger
- Default look moved with contemporary taste
Version notes
V7 was a routine default swap: better understanding and consistency. Web editing matured; Discord stopped being the only door.
Dec 2023
V6
Better prompt adherence and text.
View details
New capabilities
- V6 prompt adherence and in-image text
What got stronger
- Poster titles stopped being unreadable nine times out of ten
Version notes
V6 made “draw what I said” and “put words on the image” usable. For design workflows that beat another pretty filter.
Mar 2023
V5
Realism and anatomy.
View details
New capabilities
- V5 realism and anatomy
What got stronger
- Fewer deformed people and products
Version notes
V5 walked toward realism for commercial-photo looks. Taste debates started here: some people missed V4’s illustration feel.
Nov 2022
V4
Taste and composition jump.
View details
New capabilities
- V4 taste and composition jump
What got stronger
- Posters and concept art became usable as-is
Version notes
V4 turned Midjourney from interesting to deliverable. A lot of brand memory is still this era’s mood.
Mar 2022
Midjourney public
Discord generation community.
View details
New capabilities
- Discord bot image gen
- Community taste and prompt culture
What got stronger
- More stylized than most open models of the time
Version notes
Midjourney used Discord as gallery and social layer instead of a site-first approach. That became its community moat.