Just saw the announcement. $10/month for "Turbo Basic" and $120/month for "Turbo Pro"? For "faster" image generation? Let's cut through the marketing.
I ran a quick test on a standard prompt against the current "Fast" mode. My average generation time was around 45 seconds. Their announcement claims Turbo Basic can cut that to "under 20 seconds." Let's be generous and say it's 18 seconds.
Here's the math:
* Current $10/month "Basic" plan: ~200 GPU minutes/month on Fast mode.
* New $10/month "Turbo Basic": They don't give GPU minutes, they give "relaxed" and "turbo" time. It's a usage cap, not unlimited. You're paying the same price primarily for a speed increase.
* The value is only there if your bottleneck is absolute wall-clock time, not cost per image. For most individual users, the 45-second wait isn't a $10/month problem.
The $120 "Turbo Pro" is even wilder. It's pitched at teams, but you're now entering the territory of dedicated cloud GPU instances. For that price, you could run your own stable diffusion setup with far more control and no per-prompt censorship.
My blunt take: This is a price hike disguised as a feature tier. You're subsidizing their infrastructure scaling problems. They're selling "speed" because they can't significantly improve their core model quality fast enough to justify new pricing tiers.
If you're a solo creator, stick with Standard. If you're a team spending over $60/month on Midjourney currently, you should be looking at API-based solutions or self-hosted options. The value proposition just took a nosedive.
-- bb
-- bb
Your math on the per-image cost is the key takeaway most will miss. You're right that the value proposition shifts from 'output volume' to 'throughput time.' For hobbyists, that's rarely a critical bottleneck.
It reminds me of when cloud data warehouses introduce 'premium' query tiers. You pay a massive multiplier for faster, on-demand compute, but it only makes financial sense if your business loses real revenue during idle seconds. For most, the standard queue is fine.
I'm curious if they've introduced any new rate-limiting or quality-of-service rules behind the scenes. Sometimes these 'turbo' lanes just give you more consistent access to the same hardware, not necessarily a different SKU. The announcement language on 'usage caps' vs. 'GPU minutes' feels deliberately obfuscated.
Extract, transform, trust
Your point about comparing the $120 tier to dedicated cloud GPU instances is spot on. I ran some back-of-the-napkin numbers, and for that cost you could get a reserved Spot instance with comparable hardware, which shifts the cost-benefit analysis entirely to control and pipeline integration.
The pricing feels less like a new feature and more like market segmentation, extracting value from users whose time sensitivity outweighs their cost sensitivity. It's a classic tiering strategy, similar to how CRMs charge per "feature" seat. The real question is whether the underlying cost to serve in that turbo queue is 10x higher, or if it's just margin optimization.
Yeah, the CRM comparison is perfect. It's exactly like charging extra for "automated workflow seats" even though the underlying compute cost is minimal.
I think you're right about the segmentation. For most of us, that "Spot instance" math is the real alternative. But the key is the *pipeline integration* you mentioned. The value prop isn't just raw GPU, it's the zero-maintenance, no-ops, fully-managed service.
The real target for that $120 tier isn't hobbyists, it's micro-businesses or solo creators where waiting 45 seconds breaks a client demo or a live content workflow. They're paying for the guarantee, not just the speed. Still feels steep though!