Midjourney V8 Alpha dropped March 17, 2026 on alpha.midjourney.com. After months of anticipation – and some delays – the new model is available for community testing.
The headline improvements are substantial: roughly 5x faster generation, native 2K resolution, and text rendering that actually works. The catch: premium features now cost 4x more.
What’s new in V8
Speed. Generation is approximately five times faster than V7. The web interface has been upgraded to match, with new conversation mode, grid view, and sidebar settings that don’t block your workspace.
Native 2K resolution. The new –hd parameter renders images at 2K without upscaling. For prints or detail-heavy work, this is a significant quality jump.
Text rendering. Text in “quotes” now renders more reliably than any previous version. This has been a persistent weakness for Midjourney, and V8 makes real progress.
Prompt adherence. V8 follows detailed instructions more accurately. The model better understands aesthetics through personalization, style references (srefs), and moodboards.
Image coherence. Bodies, hands, and object details are more consistent. Not perfect, but noticeably improved.
The pricing reality
Here’s where it gets complicated. Jobs using –hd, –q 4, style references, or moodboards currently run 4x slower and cost 4x as much as standard generation.
Relax mode – the popular option for slower but free generation – isn’t available at launch. Midjourney says it’s building new server infrastructure and working on cheaper render modes.
For heavy users, this is a significant cost increase for premium features.
How to access
V8 is only available on alpha.midjourney.com. It’s not on the main site or Discord yet, and V8 creations won’t appear on the main website for now.
When you first start generating, V8 will prompt you to set up a personalization profile. Don’t skip this. Midjourney recommends leaning heavily into personalization and even setting –stylize to 1000 for best results.
Limitations
V8 is still a pure diffusion model. Competitors like Google’s Nano Banana and OpenAI’s native image generation use hybrid architectures that can achieve better prompt precision in some cases. In my initial testing with complex prompts – particularly ones requiring unusual subject relationships – V8 still struggles where autoregressive models succeed.
Upscaling, inpainting, and retexturing currently fall back to V6 models. Updates are coming.
Quick verdict
V8 is the fastest, most capable Midjourney model yet. The speed improvement alone changes the creative workflow – you can iterate much more rapidly. Native 2K output is excellent for anyone doing print work or needing high resolution.
The 4x cost multiplier on premium features will sting for power users. If you rely heavily on style references and moodboards, calculate the impact before committing.
For photorealistic or controlled output, start with –raw mode. For stylized work, embrace personalization fully.
