Midjourney vs API image models: when a subscription beats pay-per-image
Midjourney sells image generation as a monthly plan. Almost every other image model is sold per image (or per token) through an API. Which one is cheaper depends on how many images you make, whether you can wait for them, and whether a program or a person is doing the generating. This guide works out what a Midjourney image costs on each plan and puts that next to the official API prices on our image prices page.
The short answer
- If a person makes images by hand and makes a lot of them, Midjourney Standard ($30/month) or higher is hard to beat: its Relax mode has no image cap, and at the typical one minute per prompt even its Fast hours work out to under 1 cent per image.
- If you make a few hundred images a month or fewer, a cheap API model usually costs less than any plan.
- If software has to generate the images (an app, a batch script, a workflow tool), Midjourney is not an option. Its terms say: “You may not use automated tools to access, interact with, or generate Assets through the Services.” There is no public Midjourney API. Its 2025 post about an Enterprise API described it as something it was “starting to investigate”.
What a Midjourney plan includes
From Midjourney’s plan comparison page (monthly billing):
| Plan | Price per month | Fast GPU time | Relax mode | Annual billing |
|---|---|---|---|---|
| Basic | $10 | 3.3 h (200 min) | No | $8/month ($96/year) |
| Standard | $30 | 15 h | Unlimited images | $24/month ($288/year) |
| Pro | $60 | 30 h | Unlimited images and SD video | $48/month ($576/year) |
| Mega | $120 | 60 h | Unlimited images and SD video | $96/month ($1,152/year) |
According to the plan comparison page, extra Fast time costs $4 per hour on every plan. Pro and Mega add Stealth Mode, which keeps your images out of public view (Midjourney promises “best efforts”, and images made in shared spaces such as Discord stay visible there). A company (or an employee of a company) with more than $1,000,000 a year in revenue must be on Pro or Mega to own what it generates. The current default model is V8.2, which became the default on July 24, 2026.
Turning GPU hours into a price per image
Midjourney charges GPU time, not images. Its GPU speed page says one image prompt “usually takes about one minute of GPU time”. Its table lists an SD prompt at 0.8 minutes and an HD prompt at 1.3 minutes. One prompt creates a set of four images.
Using one minute per prompt (four images):
| Plan | Prompts from Fast time | Images | Cost per prompt | Cost per image |
|---|---|---|---|---|
| Basic ($10) | about 200 | about 800 | $0.050 | $0.0125 |
| Standard ($30) | about 900 | about 3,600 | $0.033 | $0.0083 |
| Pro ($60) | about 1,800 | about 7,200 | $0.033 | $0.0083 |
| Mega ($120) | about 3,600 | about 14,400 | $0.033 | $0.0083 |
With HD prompts (1.3 minutes) the Standard plan works out to about $0.011 per image, and with SD prompts (0.8 minutes) about $0.0067. Extra Fast hours bought at $4 come to about $0.017 per image at one minute per prompt.
Two things make the real cost lower or higher than this table:
- Relax mode is free on Standard and up. Relax time is “unlimited”, but jobs wait in a queue (Midjourney gives a range of 0 to 30 minutes, depending on how much Relax you have used compared with other users). If you can wait, every image beyond your Fast hours costs nothing extra.
- You rarely keep all four. If you pick one image from each set of four, the useful number is the cost per prompt ($0.033 on Standard), not the cost per image. An API returns only the image you asked for, so compare per kept image when you choose.
Turbo mode is faster but uses twice the Fast time per image, which doubles these costs.
What API image models cost
Official prices for one image at about 1K resolution (unless stated), from our image prices page. That page was last fully checked on October 19, 2026; the Google Gemini API, xAI, Runway, Luma and MiniMax prices below also matched their official pages on October 26, 2026.
| Model | Provider | Per image | 1,000 images |
|---|---|---|---|
| MiniMax image-01 | MiniMax | $0.0035 | $3.50 |
| Recraft V4.1 Flash | Recraft | $0.007 | $7 |
| FLUX 3 Image Klein | Black Forest Labs | $0.009 | $9 |
| Imagen 4 Fast | Google (Vertex AI) | $0.02 | $20 |
| Runway Gen-4 Image Turbo | Runway | $0.02 | $20 |
| Nano Banana 2.1 | $0.0336 | $33.60 | |
| Grok Imagine Image 2.0 (1K, low) | xAI | $0.04 | $40 |
| Imagen 4 | Google (Vertex AI) | $0.04 | $40 |
| Luma Uni-1.1 (2K) | Luma | about $0.0404 | about $40.40 |
| FLUX 3 Image | Black Forest Labs | $0.048 | $48 |
| Runway Gen-4 Image (720p) | Runway | $0.05 | $50 |
| Imagen 4 Ultra | Google (Vertex AI) | $0.06 | $60 |
| Nano Banana 2 | $0.067 | $67 | |
| Recraft V4.1 Pro | Recraft | $0.21 | $210 |
OpenAI’s GPT Image models are billed per token rather than per image ($30 per million image output tokens for GPT-Image-2.5 Flare and Sunburst), so the price per picture depends on size and quality settings. See the image prices page for the token rates.
Higher resolutions cost more on most of these. Nano Banana 2.1, for example, is $0.0504 at 2K and $0.113 at 4K.
Break-even: how many images before a plan is cheaper
Divide the plan price by the API price per image to get the number of API images you could buy for the same money:
| API model (per image) | Same money as Basic ($10) | Same money as Standard ($30) | Same money as Pro ($60) |
|---|---|---|---|
| MiniMax image-01 ($0.0035) | 2,857 | 8,571 | 17,143 |
| FLUX 3 Image Klein ($0.009) | 1,111 | 3,333 | 6,667 |
| Imagen 4 Fast ($0.02) | 500 | 1,500 | 3,000 |
| Nano Banana 2.1 ($0.0336) | 298 | 893 | 1,786 |
| Imagen 4 ($0.04) | 250 | 750 | 1,500 |
| FLUX 3 Image ($0.048) | 208 | 625 | 1,250 |
| Nano Banana 2 ($0.067) | 149 | 448 | 896 |
So if you keep fewer than about 900 Nano Banana 2.1 images a month, the API costs less than Midjourney Standard. Above that, and especially if you use Relax mode, the subscription wins on price. Against the cheapest API models (MiniMax image-01, FLUX 3 Image Klein), you would need thousands of kept images a month before Standard pays off.
These are price comparisons only. The models do not produce the same images, so the right choice also depends on which output you like.
When each option makes sense
Pick a Midjourney plan when:
- You generate images by hand in the web app, often hundreds or thousands a month.
- You can wait for Relax jobs (Standard and up).
- You like Midjourney’s look enough that cheaper models are not a real substitute.
Pick a pay-per-image API when:
- Your software generates the images. Midjourney’s terms forbid automated access, and it has no public API.
- You make a few hundred images a month or fewer.
- You need a specific model feature, such as image editing, vector output (Recraft) or a specific resolution, and want to pay only for what you use.
- Your usage swings a lot from month to month.
Many people use both: a plan for exploring by hand, and an API for anything that runs in code.
Related pages
- Image generation prices: every image API we track, with resolution tiers.
- Midjourney on our apps page: plan prices and links.
- AI image and video generation prices explained.
- Subscription plans: ChatGPT, Claude, Gemini and Grok plans side by side.
Sources
- Midjourney: Comparing Midjourney Plans
- Midjourney: GPU Speed (Fast, Relax, Turbo)
- Midjourney: Getting Started Guide
- Midjourney: Version
- Midjourney: Terms of Service
- Midjourney: Enterprise API Survey
- Gemini API pricing (Nano Banana 2.1, Nano Banana 2)
- Vertex AI pricing (Imagen 4)
- xAI API pricing (Grok Imagine Image 2.0)
- xAI: Grok Imagine Image 2.0
- Black Forest Labs release notes (FLUX 3 Image, FLUX 3 Image Klein)
- Runway API pricing
- Luma API pricing
- Recraft API pricing
- MiniMax pay-as-you-go pricing
- OpenAI API pricing (GPT Image models)
- OpenAI: GPT-Image-2.5 Flare
- OpenAI: GPT-Image-2.5 Sunburst