Why pricing transparency is broken in AI image generation
Every major AI image platform obscures its true per-image cost. Midjourney charges by "fast hours" and "GPU minutes." OpenAI charges different rates for "standard" vs "HD" DALL-E 3 without clearly defining the quality difference. Vertex AI charges per "character" for Imagen 4 prompts, not per image. Fal.ai is the most transparent — it lists actual dollar prices — but even there, the cost varies by resolution tier and aspect ratio in ways that are not obvious from the main pricing page.
The result is that developers and procurement teams regularly overpay by 200–400% because they chose a premium-tier generation where a standard-tier would have produced identical output for their use case. This guide cuts through the obfuscation and gives you true cost-per-image in USD for every major model, updated monthly.
The complete 2026 pricing table (standard resolution, 1:1 aspect ratio)
| Model | Cost per image (API) | Resolution tier | Notes |
|---|---|---|---|
| FLUX 1.1 Pro (fal.ai) | $0.04 | Up to 1MP | Most transparent pricing. Scales with resolution. |
| FLUX Dev (fal.ai) | $0.025 | Up to 1MP | Open-weight model. Slightly lower quality than Pro. |
| FLUX Schnell (fal.ai) | $0.003 | Up to 1MP | Fastest FLUX. Quality trade-off noticeable at fine detail. |
| DALL-E 3 Standard (OpenAI) | $0.04 | 1024×1024 | Included in ChatGPT Plus (15/day limit). |
| DALL-E 3 HD (OpenAI) | $0.08 | 1024×1024 HD | Noticeable quality improvement for text and fine detail. |
| Imagen 4 Standard (Vertex) | $0.03–$0.05 | Up to 1024px | Priced by prompt length + output resolution. Estimate range. |
| Imagen 4 Ultra (Vertex) | $0.06–$0.10 | Up to 4096px | Premium tier. Best for luxury product and food photography. |
| Seedream 3 (fal.ai) | $0.015–$0.025 | Up to 1MP | Strong value tier. Competitive quality at lower cost. |
| Ideogram v3 API | $0.04–$0.06 | Up to 1024px | Best for text-in-image. API separate from subscription. |
| Midjourney v7 (subscription) | $0.08–$0.12 | ~1MP default | No public API. Subscription ÷ monthly images. Estimate only. |
| SD 3.5 Large (self-hosted) | $0.005–$0.015 | Up to 1MP | GPU compute cost on A100. Varies by cloud provider. |
Hidden costs: what pricing pages do not tell you
- Retry charges: Some providers deduct credits for failed generations. DALL-E 3 does not charge for safety-filter rejections, but others do. Always check the failed-generation policy before using a model in a production pipeline.
- Resolution upcharges: Most models have multiple resolution tiers. FLUX Pro at standard resolution is $0.04, but at 4K ultra-high resolution it can exceed $0.15. Know your required output resolution before estimating costs.
- API vs UI pricing gap: Midjourney's subscription is priced as a UI tool. If Midjourney ever releases a public API, the API pricing historically runs 2–5x higher than subscription-equivalent rates.
- Aspect ratio multipliers: Some providers charge more for non-square aspect ratios (wide or tall crops). FLUX on fal.ai uses a pixel count calculation, so a wide 16:9 at 1MP costs the same as a 1:1 at 1MP — but a 4MP 16:9 costs significantly more.
- Upscaling fees: Many workflows require upscaling from 1MP to 4K for print or large-format use. This is a separate API call with separate charges at most providers. Budget for a 1.5–3x cost multiplier if your workflow requires upscaling.
Real cost scenarios
Scenario A: Blogger, 60 images/month (20 posts × 3 images each)
- FLUX 1.1 Pro: 60 × $0.04 = $2.40/month
- Midjourney subscription: $10–$30/month (fixed, regardless of usage)
- Smart routing recommendation: FLUX Dev or Seedream for most blog images, Imagen 4 for food/lifestyle. Estimated cost: $1.20–$1.80/month.
Scenario B: E-commerce store, 2,000 images/month (500 SKUs × 4 angles)
- FLUX 1.1 Pro: 2,000 × $0.04 = $80/month
- Imagen 4 Ultra: 2,000 × $0.08 = $160/month
- Smart routing (FLUX for most, Imagen 4 for cosmetics/food): $55–$70/month
- Annual smart routing savings vs FLUX-only: $120–$300
Scenario C: Marketing agency, 1,200 images/month
- FLUX 1.1 Pro for everything: 1,200 × $0.04 = $48/month
- Smart routing (mix of FLUX Pro, Seedream, DALL-E 3, Ideogram): $28–$35/month
- Monthly savings: $13–$20. Annual: $156–$240.
Scenario D: Developer pipeline, 10,000 images/month
- FLUX 1.1 Pro: $400/month
- Smart routing with quality floor set at 7/10: $200–$280/month
- Annual savings: $1,440–$2,400
Smart routing savings: how it works at scale
Smart routing saves money by routing each generation to the cheapest model that clears your quality threshold. Here is the logic:
- Set a quality floor (e.g., 7/10 on the eaxy quality score).
- For each generation request, the router checks which models can reliably clear that threshold for the task type.
- Among qualifying models, it routes to the cheapest one.
- For client-facing hero images (high quality threshold), it routes to FLUX Pro or Imagen 4 Ultra.
- For internal mockups and social variants (lower threshold), it routes to FLUX Schnell or Seedream.
- Result: 30–60% average cost reduction across a mixed-use generation portfolio.
When to pay for premium vs route to budget models
- Always use premium (FLUX Pro, Imagen 4 Ultra): Client-facing hero images, luxury product photography, final campaign deliverables, print-quality outputs.
- Route to mid-tier (FLUX Dev, Seedream): Social media variants, secondary images, draft client presentations, A/B test candidates before winnowing.
- Route to budget (FLUX Schnell, SD self-hosted): Internal mockups, concept exploration, batch generation for testing prompt directions before committing to full production.
Calculate your smart routing savings — enter your monthly image volume and use case mix to see your estimated monthly spend with eaxy versus single-model pricing.