# Midjourney v7

Midjourney v7 is a text-to-image model released in alpha by [Midjourney](https://www.edgepedia.org/midjourney) on April 3, 2025, the company's first new image model in nearly a year and the first with personalization switched on by default.<sup>[1](https://docs.midjourney.com/hc/en-us/articles/32199405667853-Version)</sup><sup> • </sup><sup>[2](https://updates.midjourney.com/v7-alpha/)</sup> It arrived a week after OpenAI debuted a new image generator in ChatGPT.<sup>[3](https://techcrunch.com/2025/04/03/midjourney-releases-its-first-new-ai-image-model-in-nearly-a-year/)</sup> The Midjourney company and the broader model family are covered in separate articles.

| Key fact | Detail |
|---|---|
| Alpha release | April 3, 2025<sup>[3](https://techcrunch.com/2025/04/03/midjourney-releases-its-first-new-ai-image-model-in-nearly-a-year/)</sup> |
| Default period | June 17, 2025 to June 9, 2026<sup>[1](https://docs.midjourney.com/hc/en-us/articles/32199405667853-Version)</sup> |
| Headline features | Draft Mode, personalization by default, Omni Reference<sup>[1](https://docs.midjourney.com/hc/en-us/articles/32199405667853-Version)</sup> |
| Draft Mode economics | 10x the speed, half the cost of standard generation<sup>[2](https://updates.midjourney.com/v7-alpha/)</sup> |
| Personalization unlock | Rating around 200 images<sup>[3](https://techcrunch.com/2025/04/03/midjourney-releases-its-first-new-ai-image-model-in-nearly-a-year/)</sup> |
| Entry price | $10/month Basic plan<sup>[4](https://decrypt.co/313704/midjourney-v7-review-struggles-keep-up)</sup> |
| Independent verdict | Modest improvement over v6.1, rivals ahead in realism and precision (Decrypt, April 2025)<sup>[4](https://decrypt.co/313704/midjourney-v7-review-struggles-keep-up)</sup> |

## Release and rollout

V7 began rolling out in alpha around midnight ET on Thursday, April 3, 2025, available in Turbo and Relax modes.<sup>[2](https://updates.midjourney.com/v7-alpha/)</sup><sup> • </sup><sup>[3](https://techcrunch.com/2025/04/03/midjourney-releases-its-first-new-ai-image-model-in-nearly-a-year/)</sup> At launch, turbo jobs cost 2x a normal V6 job, while draft jobs cost half as much.<sup>[2](https://updates.midjourney.com/v7-alpha/)</sup> The model became Midjourney's default on June 17, 2025 and remained default until June 9, 2026, according to the company's version documentation; the sources do not identify what replaced it as default.<sup>[1](https://docs.midjourney.com/hc/en-us/articles/32199405667853-Version)</sup>

## Draft Mode and personalization

Draft Mode is v7's most distinctive operational feature: it renders images at 10 times the speed and half the cost of standard generation, producing lower-quality drafts that can then be enhanced or varied at full quality.<sup>[2](https://updates.midjourney.com/v7-alpha/)</sup>

V7 is also the first Midjourney model with personalization on by default. Users must unlock personalization by rating images, which the company says takes about 5 minutes; independent reporting puts the requirement at around 200 rated images to build a profile that tunes the model to individual preferences.<sup>[2](https://updates.midjourney.com/v7-alpha/)</sup><sup> • </sup><sup>[3](https://techcrunch.com/2025/04/03/midjourney-releases-its-first-new-ai-image-model-in-nearly-a-year/)</sup><sup> • </sup><sup>[5](https://www.tomsguide.com/ai/midjourney-version-7)</sup> While personalization was possible on earlier models, this was the first time it was switched on by default.<sup>[5](https://www.tomsguide.com/ai/midjourney-version-7)</sup> The sources do not document a '--p off' or raw-mode mechanism for disabling it beyond the option to revert to V6.1 in settings.<sup>[6](https://updates.midjourney.com/v7-is-now-the-default-model/)</sup>

## Architecture and training as disclosed

Midjourney's only technical statement about v7 came from CEO David Holz, who described it as a "totally different architecture" in a post on X, with no further disclosure of parameters, training data, or training method.<sup>[3](https://techcrunch.com/2025/04/03/midjourney-releases-its-first-new-ai-image-model-in-nearly-a-year/)</sup>

## By the numbers

The vendor and independent sources give these figures:

- **Speed and cost:** Draft Mode renders at 10x speed and half cost; turbo jobs cost 2x a normal V6 job.<sup>[2](https://updates.midjourney.com/v7-alpha/)</sup>
- **Personalization preference:** Midjourney reports improved personalization profiles are preferred by 85% of users.<sup>[6](https://updates.midjourney.com/v7-is-now-the-default-model/)</sup>
- **User base:** Midjourney still commanded a Discord-driven user base of some 20 million people at v7's release.<sup>[4](https://decrypt.co/313704/midjourney-v7-review-struggles-keep-up)</sup>
- **Pricing:** Subscriptions start at $10/month; one review source lists tiers from $10/month Basic (about 3.3 fast GPU hours/month) to $120/month Mega, metered in GPU hours, with generation through the web app and Discord rather than a public API. This rests on a single review source and is not confirmed by primary documentation.<sup>[4](https://decrypt.co/313704/midjourney-v7-review-struggles-keep-up)</sup><sup> • </sup><sup>[8](https://look4ai.net/tool/midjourney-v7)</sup>

## How it compares with v6.1 and rivals

Midjourney claims v7 offers higher image quality, better prompt understanding, and improved image coherence versus v6.1.<sup>[6](https://updates.midjourney.com/v7-is-now-the-default-model/)</sup> Independent testing by Decrypt in April 2025 found a modest improvement over v6.1 across realism, prompt adherence, anatomy, and artistic style, but "not the kind of upgrade that reclaims the crown."<sup>[4](https://decrypt.co/313704/midjourney-v7-review-struggles-keep-up)</sup> Decrypt judged that newer tools, including OpenAI's GPT-4o, Reve, and Ideogram 3.0, had overtaken Midjourney in realism, precision, and functionality by that date; in its tests, ChatGPT at $20/month showed better prompt adherence and spatial awareness than Midjourney's $10/month entry tier.<sup>[4](https://decrypt.co/313704/midjourney-v7-review-struggles-keep-up)</sup>

Only qualitative hands-on testing comparing v7 against Flux, Seedream, Imagen, or [GPT Image](https://www.edgepedia.org/gpt-image) exists.<sup>[4](https://decrypt.co/313704/midjourney-v7-review-struggles-keep-up)</sup>

## Reception and the 'lateral move' debate

Reception was polarized. Critics called the improvements incremental, describing v7 as "V6.2" rather than a true leap, citing anatomy and text-generation issues, and the loss of previous features like `--cref`.<sup>[7](https://theaitrack.com/midjourney-v7-launch/)</sup> Influencers including Ethan Mollick, Javi Lopez, and David Shapiro posted critical feedback, noting difficulty reproducing v6-quality results with identical prompts.<sup>[7](https://theaitrack.com/midjourney-v7-launch/)</sup> On the other side, power users praised image quality, voice input, and Draft Mode.<sup>[7](https://theaitrack.com/midjourney-v7-launch/)</sup> Reviewers also disagree on v7's overall standing: Look4AI calls it "the category benchmark" for pure image aesthetics, while Decrypt says the former industry standard "struggles to keep up." Look4AI's review finds v7 still struggles with legible in-image text and precise layout control.<sup>[4](https://decrypt.co/313704/midjourney-v7-review-struggles-keep-up)</sup><sup> • </sup><sup>[8](https://look4ai.net/tool/midjourney-v7)</sup>

## What changed between April 2025 and September 2026

At alpha launch, upscaling, editing, and retexture fell back to V6 models, with updates promised; [TechCrunch](https://www.edgepedia.org/techcrunch) reported these were expected possibly within two months.<sup>[2](https://updates.midjourney.com/v7-alpha/)</sup><sup> • </sup><sup>[3](https://techcrunch.com/2025/04/03/midjourney-releases-its-first-new-ai-image-model-in-nearly-a-year/)</sup> The biggest announced roadmap feature was a new v7 character and object reference, which shipped as Omni-reference (`--oref`), letting users put consistent characters and objects into scenes, alongside an update to Style References.<sup>[2](https://updates.midjourney.com/v7-alpha/)</sup><sup> • </sup><sup>[6](https://updates.midjourney.com/v7-is-now-the-default-model/)</sup> After months of testing and these additions, Midjourney made v7 the default model in mid-June 2025, with users able to revert to v6.1 via web settings or `/settings` on Discord.<sup>[6](https://updates.midjourney.com/v7-is-now-the-default-model/)</sup> V7 ceased to be the default on June 9, 2026; the sources do not name its successor or confirm a "v7.1" release.<sup>[1](https://docs.midjourney.com/hc/en-us/articles/32199405667853-Version)</sup>

## References

1. [Version – Midjourney (official docs)](https://docs.midjourney.com/hc/en-us/articles/32199405667853-Version)
2. [V7 Alpha (Midjourney official announcement)](https://updates.midjourney.com/v7-alpha/)
3. [Midjourney releases V7, its first new AI image model in nearly a year (TechCrunch)](https://techcrunch.com/2025/04/03/midjourney-releases-its-first-new-ai-image-model-in-nearly-a-year/)
4. [Midjourney v7 Review: The Former Industry Standard Struggles to Keep Up (Decrypt)](https://decrypt.co/313704/midjourney-v7-review-struggles-keep-up)
5. [Midjourney V7 has arrived: Here's everything you need to know (Tom's Guide)](https://www.tomsguide.com/ai/midjourney-version-7)
6. [V7 is now the default model! (Midjourney official update)](https://updates.midjourney.com/v7-is-now-the-default-model/)
7. [Midjourney V7: Midjourney's First New Model in Nearly a Year (The AI Track)](https://theaitrack.com/midjourney-v7-launch/)
8. [Midjourney V7 review: Class-leading AI image aesthetics with a fast, cheap Draft Mode (Look4AI)](https://look4ai.net/tool/midjourney-v7)

---
*Topic: Encyclopedia › Technology and the built world › Computing and digital systems › Modern AI: foundation models, generative AI, and the AI industry › Model families and named models › Image generation models*

*Initially written Sep 17, 2026 · Edited: Oct 11, 2026; Oct 15, 2026*

*Copyright 2026 EdgeChat AI, a subsidiary of Biostate AI.*

License: Edgepedia Community License 1.0, https://www.edgepedia.org/edgepedia/license
