Live now

GPT Image 2 is live on fal.ai

April 21, 2026Published on launch day5 min read

GPT Image 2 is live on fal as of 12PM PT today. Two endpoints ship together:

- openai/gpt-image-2 for text-to-image
- openai/gpt-image-2/edit for reference image editing

Open to everyone on day one. If you have a FAL_KEY, you can call it right now. The playground on this site runs against the new model by default so you can iterate without writing code.

The headline in one sentence

This is the first OpenAI image model where you can ship typography-critical output without a human review in the loop. Near-perfect glyph accuracy, neutral color, native 4K, and single-pass inference that lands under 3 seconds at medium quality. Everything below is about what to do with that.

Six things that shipped today

Near-perfect text rendering

Over 99 percent glyph accuracy on English, CJK scripts finally reliable, multi-line typography that ships to production without a human review gate.

Photoreal faces and skin

Over 70 percent of blind test participants misclassify GPT Image 2 portraits as real photographs. Skin micro detail, catchlights, and hands all land cleaner.

Neutral color, no yellow cast

The amber wash in gpt-image-1.5 daylight scenes is gone. Daylight is daylight, gray is gray, brand colors match the hex you asked for.

Native 4K and 16:9

Three new sizes: 1920x1080, 2560x1440, and native 3840x2160. No more upscaling a 1792x1024 in post for a landing page banner.

Single-pass inference

The old planner turn is gone. Expect p50 near 3 seconds at medium quality. The whole render feel changes.

Edit fidelity

The /edit endpoint gains input_fidelity so you can swap a background, restage a product, or change typography while the subject stays locked.

Migrate in one env var

Pin the model string behind a constant so the switch is one environment change instead of a code edit. Ship once, flip the var, let the rollout do the rest.

import { fal } from "@fal-ai/client";

fal.config({ credentials: process.env.FAL_KEY });

const IMAGE_MODEL = process.env.FAL_IMAGE_MODEL ?? "openai/gpt-image-2";

export async function render(prompt: string) {
  return fal.subscribe(IMAGE_MODEL, {
    input: {
      prompt,
      image_size: "1024x1024",
      quality: "high",
      num_images: 1,
      output_format: "png",
    },
  });
}

Set FAL_IMAGE_MODEL=openai/gpt-image-2 in your environment, redeploy, and the whole pipeline runs against the new model. No code change, one env var. Roll back to fal-ai/gpt-image-1.5 just as fast if you need to.

Pricing

Per-image pricing across the three quality tiers. The live rate card sits on the openai/gpt-image-2 model page; check there before quoting a customer.

Resolution	Low	Medium	High
1024x768	$0.01	$0.04	$0.15
1024x1024	$0.01	$0.06	$0.22
1024x1536	$0.01	$0.05	$0.17
1920x1080	$0.01	$0.04	$0.16
2560x1440	$0.01	$0.06	$0.23
3840x2160	$0.02	$0.11	$0.41

Who should switch on day one

If your product renders text on images (UI mockups, packaging, infographics, posters, YouTube thumbnails, book covers), flip the flag today. The typography gain removes the human review cost that dominated your editorial pipeline.

If your product is photoreal without typography (product shots, real estate hero stills, food editorial), run a quick bench before switching. The 1.5 output was already good; 2.0 is better, but the delta is smaller.

If your product is stylized artistic work (illustration, moodboard, painterly marketing), run both in parallel for a week and pick per brief. GPT Image 2 wins on most prompts; Flux 2 Pro and other stylized siblings still win on a few.

Migration checklist

Pin the model string behind an env var like FAL_IMAGE_MODEL
Prepare a golden test suite of 10 to 30 prompts that define your product
Add a feature flag that routes a subset of traffic to openai/gpt-image-2 first
Run the golden suite side by side against gpt-image-1.5 and diff the outputs
Bump production traffic to 100 percent once the diff review is clean
Update your API docs and runbooks to reference openai/gpt-image-2

What this site runs on today

Every tool on gptimage2prompts.com defaults to openai/gpt-image-2. The playground, image editor, style reference, and image to image flows all call the new model out of the box and let you flip back to gpt-image-1.5 from the Model dropdown for a side-by-side. Your fal API key keeps working; nothing else changes for you.

The prompt library is rerendering every featured prompt on the new model; the gpt-image-1.5 reference renders stay on file so you can see the quality delta side by side.

The endpoint is live

openai/gpt-image-2 went live on fal at 12PM PT today. Paste your fal key in settings and the playground calls it on the next request.

Add your fal key Open the playground