Every major image API, side by side: OpenAI's GPT-Image models, Google's Gemini image models, ByteDance Seedream, Black Forest Labs FLUX, Recraft and Ideogram. Some bill a flat fee per image, others bill tokens or scale by resolution, so the headline price only tells part of the story.
Flat per-image APIs are the easiest to budget: Seedream 4.5 and Recraft V3 each list $0.04 / image (raster output), and the open-weights FLUX Schnell is listed at $0.003 / image. Google's Gemini 3.1 Flash Image scales with resolution ($0.045–0.151 / image), while OpenAI's GPT-Image models bill per token, so quality and size settings change what each image costs.
| Model | Provider | Price | Notes |
|---|---|---|---|
| GPT-Image-2.5 Flare | OpenAI | ≈$0.064 / image | New flagship (Sep 8, 2026), default fast variant; same per-token rates as GPT-Image-2 ($8 image in / $30 image out per 1M) with higher quality & lower latency |
| GPT-Image-2.5 Sunburst | OpenAI | ≈$0.064 / image | Premium editing/control variant (Sep 8, 2026); shares Flare's published per-token price |
| GPT-Image-2 | OpenAI | ≈$0.053 / image | Superseded by GPT-Image-2.5 Flare/Sunburst (Sep 8, 2026) as OpenAI's flagship, but still offered; photoreal edits; medium quality at 1024×1024 shown (low ≈$0.006, high ≈$0.211) |
| GPT-Image-1.5 | OpenAI | ≈$0.034 / image | Being retired Dec 1, 2026 (along with GPT-Image-1-mini) — migrate to GPT-Image-2.5 |
| Gemini 3.1 Flash Image | $0.045–0.151 / image | Tiered by resolution: $0.045 @512px, $0.067 @1024px, $0.151 @4K | |
| Gemini 3 Pro Image | $0.134–0.24 / image | "Nano Banana Pro"; premium tier above Flash Image, up to 4K | |
| Seedream 4.5 | ByteDance | $0.04 / image | Strong realism & spatial understanding |
| FLUX 1.1 Pro | BFL | $0.04 / image | Open-ecosystem favorite |
| Recraft V3 | Recraft | $0.04 / image | Design, vector styles & long text |
| Model | Provider | Price | Notes |
|---|---|---|---|
| Ideogram V3 Turbo | Ideogram | $0.03 / image | Fast; best-in-class text rendering |
| GPT-Image-1-mini | OpenAI | ≈$0.011 / image | Cheap OpenAI tier; being retired Dec 1, 2026 along with GPT-Image-1.5 |
| FLUX Schnell | BFL | $0.003 / image | Cheapest solid quality; open weights |
Three billing models are in use. ByteDance (Seedream on BytePlus ModelArk) and Recraft charge a flat fee per output image. Google prices Gemini image models by output resolution: Gemini 3.1 Flash Image spans $0.045–0.151 / image from 512px up to 4K, and a 4K image costs more than twice a 1024px one, so a pipeline that defaults to 4K pays that premium on every prompt. OpenAI bills GPT-Image models by tokens (image input, image output and text input), and its own docs warn that equal token rates don't mean equal cost per image: the quality setting and size decide how many output tokens an image uses. Read the usage field on real responses before you commit to a per-image budget.
OpenAI recommends the GPT-Image-2.5 models for new integrations. Flare is the fast everyday generator; Sunburst is aimed at workflows where editing precision matters most, and both add xhigh and max quality levels above the high ceiling of earlier models. GPT-Image-1.5 and GPT-Image-1-mini shut down on December 1, 2026, and GPT-Image-1 already shut down on October 23, 2026, so don't start new work on them. Two cost details worth knowing: cached image input only applies through the Responses API image tool, not direct /v1/images/edits calls, and each streamed partial image adds 100 output tokens.
Google's Batch tier halves Gemini image prices, which suits catalogue or thumbnail jobs that don't need an answer in seconds; neither Gemini image model has a free API tier. On BytePlus, input images are free for Seedream 4.5 and images blocked by content moderation are not billed. Recraft charges twice its raster rate for vector (SVG) output, so generate raster drafts and only vectorize finals. FLUX Schnell's weights are Apache 2.0 and cleared for commercial use, so at high volume self-hosting it on your own GPUs is a realistic alternative to any per-image fee.
Several providers have shipped newer image models alongside the ones above. Black Forest Labs now labels FLUX 1.1 as previous generation and prices its FLUX.2 family by megapixel; BytePlus lists Seedream 5.0 pro, flash and lite; Recraft lists V4 at the same base raster rate as V3; and Ideogram's API now lists versions 4.0 and 4.5 next to 3.0. Check those pages if you need the latest model rather than the cheapest stable one.
Of the models tracked here, FLUX Schnell is the cheapest at $0.003 / image, and because its weights are open (Apache 2.0) you can also run it yourself. GPT-Image-1-mini (≈$0.011 / image) comes next, but it's token-billed and shuts down on December 1, 2026; the next flat per-image price is Ideogram V3 Turbo at $0.03 / image.
Gemini 3.1 Flash Image is $0.045–0.151 / image depending on resolution, and Gemini 3 Pro Image (Nano Banana Pro) is $0.134–0.24 / image. Batch requests are billed at half those rates.
The per-image figures for OpenAI models are estimates: OpenAI doesn't publish fixed per-image prices for its current GPT-Image models, which are billed per token, and the number of output tokens depends on the quality setting and image size, plus input tokens for any reference images and the prompt. Higher quality levels and larger sizes use more tokens.
OpenAI's deprecation schedule lists December 1, 2026 for both, with GPT-Image-2.5 Sunburst or Flare as the replacement.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.