OpenAI: GPT-5.4 Image 2

fileimagetext imagetext 272K context

1 provider
2 test samples

What we tested, and what came back

Each capability below is a fixed experiment: the same prompt and the same input images for every model we test. That's what makes these outputs comparable — and it's why the same bicycle shows up on every image model's page.

Each of these is a fixed experiment — the same prompt and the same input images for every model we test, so the outputs are directly comparable.

Renders a data-faithful infographic ✓ passed
We asked

“Create a clean infographic: a vertical bar chart titled 'HIFIBOTS USERS' with exactly three labeled bars: Jan = 10, Feb = 25, Mar = 40. Show each label and value as legible text.”

We gave it

Nothing — prompt only

It returned
a vision model was asked a yes/no question about the output 108s $0.22668 provider_reported
Replaces a target object in place (prompt-based inpainting) ✓ passed
We asked

“In this image, replace the red square with a yellow star. Keep the blue circle exactly as it is.”

We gave it 1 reference image
Input fixture shapes.png
shapes.png
It returned
a vision model was asked a yes/no question about the output 108s $0.22604 provider_reported
Stated by the provider — not tested by us
✓ also understands images✓ multi reference✓ reference image✓ style transfer✓ text in image✓ text to image✗ aspect ratio control✗ background removal✗ transparency✗ upscaling
Specs
aspect ratios
1024x1024, 1536x1024, 1024x1536, 2048x2048, 2048x1152, 3840x2160, 2160x3840, custom <=3840px edge, multiples of 16px, AR<=3:1
max resolution
4K: 3840x2160 (max edge 3840px) — documented for underlying gpt-image-2
watermark
C2PA + SynthID per OpenAI, explicitly incl. gpt-image-2-era output
price per image
Per token via OpenRouter: $8/1M in, $15/1M out; image $30/1M image tokens. OpenAI examples: 1024 low ~$0.006, medium ~$0.053, high ~$0.211
speed tier
slow
latency t2i s
166.2
latency i2i s
108.8
latency s
108.20043635368347

Sources: openrouter.ai · openrouter.ai · openrouter.ai · developers.openai.com

Where to run it

The same model, priced by each provider that serves it. Cheapest first — the card's headline price is the best available anywhere, which is the wrong number for choosing where to actually send your traffic.

ProviderPriceContext
openrouter $8 / $15 per 1M272KGet a key ↗
See all openai models we've tested →

How we test

Every claim is tagged by how we know it. Tested means we ran the model against a fixed prompt and checked the output — the images above are those runs. Researched means it's documented by the provider and cited. Nothing here is inferred.