Free tool · Table
AI Image Generation Cost (2026)
How do AI image generator prices compare? Subscriptions charge a flat fee each month: Adobe Firefly lists individual plans from $9.99 to $199.99 a month, Midjourney from $10 to $120, Leonardo from $12 to $60 and Ideogram from $20 to $60. OpenAI's image API has no monthly fee and bills by the token: the per-image table in its image-generation guide converts output tokens into a cost at three standard sizes, from $0.005, the lowest figure in that table, which lists it for both gpt-image-1-mini Low at 1024×1024 and gpt-image-2 Low at 1024×1536 and 1536×1024 (gpt-image-2 Low is $0.006 at 1024×1024). The prompt's tokens are billed on top, and the guide's calculator shows lower output costs for gpt-image-2 at some smaller custom sizes. For new integrations OpenAI recommends its GPT Image 2.5 models, billed at the same image token rates as gpt-image-2; the guide's calculator estimates their output cost per image, and its per-image table covers only the earlier models. For scale, the $10 Midjourney lists for Basic covers the output of 2,000 images at $0.005, the lowest figure in OpenAI's per-image table, before prompt tokens. Google discontinued its Imagen 4 models on 30 June 2026, the date its Google Cloud model page gives, and says Imagen is shut down in the Gemini API; it points new work to its Gemini image models, such as Gemini 2.5 Flash Image. The Imagen prices in the table are what Google's price list still shows, for models you can no longer build on. Self-hosting an open-weight model such as FLUX.1-schnell, which Black Forest Labs releases under Apache 2.0, has no per-image fee; you pay for your own GPU.
What does it actually cost to make AI images? Compare flat subscriptions (Midjourney, Ideogram, Leonardo, Firefly), pay-as-you-go APIs (OpenAI gpt-image, billed by the token), and free self-hosting (Stable Diffusion, FLUX), with commercial-use rights. Prices read between June 2026 and 31 October 2026.
All. 30 of 30 plans listed.
- SubscriptionMidjourneyBasic$10/mo
~3.3 fast GPU hrs ≈ ~200 images; no relax mode
- Commercial: Yes (companies >$1M rev need Pro/Mega Stealth)
Best-in-class aesthetic quality
- SubscriptionMidjourneyStandard$30/mo
15 fast hrs ≈ ~900 images + UNLIMITED relax
- Commercial: Yes
Unlimited relax mode, the best value tier
- SubscriptionMidjourneyPro$60/mo
30 fast hrs ≈ ~1,800 + unlimited relax
- Commercial: Yes
Adds Stealth (private generations)
- SubscriptionMidjourneyMega$120/mo
60 fast hrs ≈ ~3,600 + relax + Stealth
- Commercial: Yes
Highest fast-hour allotment
- SubscriptionIdeogramFree$0
Slow credits weekly, for eligible accounts
- Commercial: Limited (paid grants rights)
Best in-image text/typography
- SubscriptionIdeogramPlus$20/mo
1,000 priority credits/mo + unlimited slow
- Commercial: Yes
Cheapest paid entry; $15/mo billed annually
- SubscriptionIdeogramPro$60/mo
3,500 priority credits/mo + unlimited slow
- Commercial: Yes
High-volume priority credits; $42/mo billed annually
- SubscriptionLeonardo.AiFree$0
150 fast tokens/day, no rollover
- Commercial: Not stated for the free plan
Image and video generation included on the free tier
- SubscriptionLeonardo.AiEssential$12/mo
8,500 fast tokens/mo + 25,500 rollover
- Commercial: Yes
10 personal AI models; $10/mo billed annually
- SubscriptionLeonardo.AiPremium$30/mo
25,000 fast tokens/mo + 75,000 rollover
- Commercial: Yes
Unlimited relaxed image generation; $24/mo billed annually
- SubscriptionLeonardo.AiUltimate$60/mo
60,000 fast tokens/mo + 180,000 rollover
- Commercial: Yes
Adds unlimited relaxed video; $48/mo billed annually
- SubscriptionAdobe FireflyFree$0
Free daily generations (limits refresh daily; no credit count published)
- Commercial: Not stated for the free plan
A free plan with limited daily generations across image, video and audio
- SubscriptionAdobe FireflyStandard$9.99/mo
2,000 generative credits/mo + unlimited standard image features
- Commercial: Yes (commercially safe)
Partner models (Google, OpenAI, Flux) on every paid plan; up to 20 five-second videos
- SubscriptionAdobe FireflyPro$19.99/mo
4,000 generative credits/mo + unlimited standard image features
- Commercial: Yes (commercially safe)
Adds Photoshop on web and mobile and Adobe Express Premium; up to 40 five-second videos
- SubscriptionAdobe FireflyPro Plus$49.99/mo
10,000 generative credits/mo + unlimited standard image features
- Commercial: Yes (commercially safe)
Up to 100 five-second videos
- SubscriptionAdobe FireflyPremium$199.99/mo
50,000 generative credits/mo + unlimited standard image features
- Commercial: Yes (commercially safe)
Unlimited Firefly Video Model generations
- Pay-as-you-go APIOpenAIgpt-image-2 · Low (1024²)$0.006/image
Per image (OpenAI's per-token billing at a standard size); batch is half price
- Commercial: Yes (you own outputs)
Cheapest gpt-image-2 tier; 1024x1536 is $0.005
- Pay-as-you-go APIOpenAIgpt-image-2 · Medium$0.053/image
Per image (OpenAI's per-token billing at a standard size); batch is half price
- Commercial: Yes
Non-square costs less: $0.041 at 1024x1536
- Pay-as-you-go APIOpenAIgpt-image-2 · High$0.211/image
Per image (OpenAI's per-token billing at a standard size); batch is half price
- Commercial: Yes
Top gpt-image-2 quality; $0.165 at 1024x1536
- Pay-as-you-go APIOpenAIgpt-image-1 · Low (1024²)$0.011/image
Per image (OpenAI's per-token billing at a standard size); batch is half price
- Commercial: Yes (you own outputs)
Undercut by gpt-image-2 Low at $0.006
- Pay-as-you-go APIOpenAIgpt-image-1 · Medium$0.042/image
Per image (OpenAI's per-token billing at a standard size); batch is half price
- Commercial: Yes
Undercut by gpt-image-1.5 Medium at $0.034
- Pay-as-you-go APIOpenAIgpt-image-1 · High$0.167/image
Per image; larger sizes cost more ($0.25)
- Commercial: Yes
Highest fidelity of gpt-image-1
- Pay-as-you-go APIOpenAIgpt-image family · range$0.005–0.25/image
mini / 1 / 1.5 / 2 across quality+size
- Commercial: Yes
$0.005 floor is a TIE: mini Low and gpt-image-2 Low
- Pay-as-you-go APIDiscontinued 30 June 2026Google ImagenImagen 4 Fast$0.02/image
Per image on Google Cloud; Google lists Imagen 4 as discontinued 2026-06-30
- Commercial: Yes
Cheapest Imagen 4 tier; speed-optimized
- Pay-as-you-go APIDiscontinued 30 June 2026Google ImagenImagen 4 Standard$0.04/image
Per image on Google Cloud; discontinued 2026-06-30
- Commercial: Yes
Balanced default quality
- Pay-as-you-go APIDiscontinued 30 June 2026Google ImagenImagen 4 Ultra$0.06/image
Per image on Google Cloud; discontinued 2026-06-30
- Commercial: Yes
Highest-fidelity Imagen tier
- Self-hostStable DiffusionSD 1.5 / SDXL (local)free
Unlimited local; ~8–16 GB VRAM
- Commercial: Yes (OpenRAIL-M, no rev cap)
Zero per-image cost + huge LoRA ecosystem
- Self-hostStable DiffusionSD 3 / 3.5 (local)free <$1M rev
Free to self-host under the revenue gate
- Commercial: Free under $1M/yr org rev; else paid license
Watch the $1M licensing gate
- Self-hostFLUX (Black Forest Labs)FLUX.1-schnell (local)free
Unlimited local; fast distilled model
- Commercial: Yes (Apache 2.0, unrestricted)
Open + commercially free + fast, a great default
- Self-hostFLUX (Black Forest Labs)FLUX.1-dev (local)free (non-commercial)
Unlimited local; ~16 GB VRAM (8 with FP8)
- Commercial: Model is non-commercial (FLUX.1 [dev] Non-Commercial License); commercial self-host needs a BFL license
Top open-weight quality (mind the license)
Verified June 2026, except Leonardo.Ai and Stable Diffusion, re-read 2026-10-31. Subscription image counts (esp. Midjourney) are ESTIMATES. Midjourney sells fast GPU time (~1 min/image), not a fixed count; Ideogram/Leonardo meter in credits/tokens where one image costs a variable amount. Pay-as-you-go API and subscription prices aren't directly comparable without an assumed monthly volume. OpenAI bills its GPT Image models by the token: the per-image figures are OpenAI's conversion of output tokens at three standard sizes, prompt tokens are extra, and for new integrations it recommends GPT Image 2.5, billed at the same image token rates as gpt-image-2. Google discontinued Imagen 4 on 30 June 2026, per its Google Cloud model page, and says Imagen is shut down in the Gemini API; the Imagen prices above are what Google's price list still shows, and Google points new work to its Gemini image models.Self-host “free” = $0 software only; you pay for the GPU, and commercial rights differ by model (FLUX.1-dev is non-commercial without a license). Confirm current pricing with each vendor.
Midjourney vs Leonardo vs Ideogram vs Firefly vs OpenAI pricing
The individual paid plans in our table, in US dollars, as each publisher lists them.
- Adobe FireflySubscription, per month
- Standard$9.99
- Pro$19.99
- Pro Plus$49.99
- Premium$199.99
- MidjourneySubscription, per month
- Basic$10
- Standard$30
- Pro$60
- Mega$120
- LeonardoSubscription, per month
- Essential$12
- Premium$30
- Ultimate$60
- IdeogramSubscription, per month
- Plus$20
- Pro$60
- OpenAIAPI, output cost per image at 1024×1024; prompt tokens extra
- gpt-image-2 Low$0.006
- gpt-image-2 Medium$0.053
- gpt-image-2 High$0.211
- gpt-image-1 Low$0.011
- gpt-image-1 Medium$0.042
- gpt-image-1 High$0.167
Sources: Midjourney's plan comparison in its help centre, Leonardo's pricing cards, Ideogram's pricing page, Adobe's Firefly plans page (US prices) and OpenAI's image-generation guide, whose per-image figures convert output tokens only. Google Imagen stays in the table above but is not quoted here. Google discontinued its Imagen 4 models on 30 June 2026, the date its Google Cloud model page gives, and says Imagen is shut down in the Gemini API; it points new work to its Gemini image models, such as Gemini 2.5 Flash Image. The Imagen prices in the table are what Google's price list still shows, for models you can no longer build on.
Want the verdict, not just the table?
The table shows every price; the guide adds the opinionated pick, which to choose for quality, for building into an app, or for free self-host.
Frequently asked
How do AI image generator prices compare?
Subscriptions charge a flat fee each month: Adobe Firefly lists individual plans from $9.99 to $199.99 a month, Midjourney from $10 to $120, Leonardo from $12 to $60 and Ideogram from $20 to $60. OpenAI's image API has no monthly fee and bills by the token: the per-image table in its image-generation guide converts output tokens into a cost at three standard sizes, from $0.005, the lowest figure in that table, which lists it for both gpt-image-1-mini Low at 1024×1024 and gpt-image-2 Low at 1024×1536 and 1536×1024 (gpt-image-2 Low is $0.006 at 1024×1024). The prompt's tokens are billed on top, and the guide's calculator shows lower output costs for gpt-image-2 at some smaller custom sizes. For new integrations OpenAI recommends its GPT Image 2.5 models, billed at the same image token rates as gpt-image-2; the guide's calculator estimates their output cost per image, and its per-image table covers only the earlier models. For scale, the $10 Midjourney lists for Basic covers the output of 2,000 images at $0.005, the lowest figure in OpenAI's per-image table, before prompt tokens. Google discontinued its Imagen 4 models on 30 June 2026, the date its Google Cloud model page gives, and says Imagen is shut down in the Gemini API; it points new work to its Gemini image models, such as Gemini 2.5 Flash Image. The Imagen prices in the table are what Google's price list still shows, for models you can no longer build on. Self-hosting an open-weight model such as FLUX.1-schnell, which Black Forest Labs releases under Apache 2.0, has no per-image fee; you pay for your own GPU.
Subscription vs pay-as-you-go API: which is cheaper for AI images?
It depends entirely on volume. Pay-as-you-go APIs (OpenAI gpt-image, billed by the token, whose per-image table puts a standard-size image's output from $0.005, before prompt tokens) win for low or spiky usage because you pay nothing when idle. Flat subscriptions (Adobe Firefly $9.99–$199.99, Midjourney $10–$120, Leonardo $12–$60 and Ideogram $20–$60) win for steady high-volume creative work: Midjourney Basic at $10 for ~200 images is ~5¢/image, and on its unlimited relax mode the effective per-image cost approaches zero, which no pay-as-you-go API can match.
What's the absolute cheapest way to generate AI images?
Self-hosting an open-weight model (Stable Diffusion SD1.5/SDXL, or FLUX.1-schnell) is effectively free per image; you only pay for your own GPU/electricity or a rented cloud GPU. Of the hosted APIs in our table, the lowest per-image figure comes from OpenAI's per-image table: $0.005, the lowest figure in that table, which lists it for both gpt-image-1-mini Low at 1024×1024 and gpt-image-2 Low at 1024×1536 and 1536×1024 (gpt-image-2 Low is $0.006 at 1024×1024). OpenAI bills by the token, so the prompt's tokens are extra, and its Batch token rates are half its standard ones. Google Imagen 4 is not an option: Google discontinued it on 30 June 2026.
How many images does a Midjourney plan actually give me?
Midjourney sells fast GPU TIME, not a fixed image count. A typical still image takes ~1 GPU-minute, so Basic's 3.3 fast hours ≈ ~200 images, Standard's 15 hours ≈ ~900, Pro's 30 ≈ ~1,800, Mega's 60 ≈ ~3,600. These are estimates that shift with model, upscales and settings, and Standard/Pro/Mega also include unlimited (slower) relax-mode generations, so the practical ceiling is far higher.
Which AI image services are safe for commercial use?
Paid Midjourney, OpenAI gpt-image, paid Ideogram, paid Leonardo, and Adobe Firefly all grant commercial rights, and Adobe calls Firefly's models commercially safe, with IP indemnification for eligible customers on its enterprise plans only. For self-host: SD1.5/SDXL and FLUX.1-schnell (Apache 2.0) are commercial-free; SD3/3.5 is free commercially only under $1M/yr org revenue; FLUX.1-dev needs a paid Black Forest Labs license for commercial self-hosting. Free tiers (Leonardo, Ideogram) are generally personal-use only.
Why do Ideogram and Leonardo show 'credits' or 'tokens' instead of images?
Both meter usage in an abstract currency where one image costs a variable number of units depending on model and quality. On Ideogram, a top-quality image costs ~6 credits while a turbo image costs a fraction of one, so a 1,000-credit plan could yield anywhere from ~167 to thousands of images. Leonardo tokens work the same way, and Ideogram priority credits expire monthly with no rollover.