If you are checking the current public OpenAI image API price for gpt-image-2, there is no standalone official gpt-image-2 pricing row to match against. As of April 17, 2026, the public billing route still maps to gpt-image-1.5 and GPT Image Latest: the main pricing page shows token billing, while the image generation guide lists per-image output pricing.
You will often see official gpt-image-1.5 pricing next to cheaper gpt-image-2 offers. The first one is the public OpenAI billing baseline. The cheaper pages usually sell a third-party access route, relay route, or reseller contract. Separate those two layers first, then decide whether you are validating OpenAI's own bill or just looking for lower-cost access.
Quick answer
| Question | Direct answer |
|---|---|
Does OpenAI have a standalone public gpt-image-2 pricing row? | No. The current public official pricing still maps to gpt-image-1.5 / GPT Image Latest. |
| Why does official pricing look like more than one number? | Because OpenAI exposes both token billing and per-image output pricing. |
What is the relationship between GPT Image Latest and gpt-image-1.5? | In the public image generation docs, gpt-image-1.5 is the latest and most advanced current public GPT Image model, and the image guide says GPT Image Latest uses the same documented per-image output pricing. |
Why do some gpt-image-2 pages look cheaper? | They are usually quoting a third-party access price, not the same OpenAI official pricing table. |
| Which official page should I trust first? | Start with OpenAI's API Pricing page for token billing, then use the image generation guide for per-image output pricing. |
Why some pages say gpt-image-2 while official pricing says gpt-image-1.5
You may see gpt-image-2, gpt-image-1.5, and GPT Image Latest in the same conversation. The confusing part is usually not the dollar figure. It is the name mapping. On third-party pages and in discussion, gpt-image-2 often gets used as a loose label for a newer GPT Image route or a cheaper compatible access option. That does not mean OpenAI's public pricing pages use the same row name.
As of April 17, 2026, OpenAI's public image-generation documentation names gpt-image-1.5, gpt-image-1, and gpt-image-1-mini, and describes gpt-image-1.5 as the latest and most advanced current public GPT Image model. So when you are validating official pricing, docs, or billing truth, map back to the current public name instead of treating gpt-image-2 as a public official pricing row.
The practical rule is straightforward. gpt-image-2 can stay as the entry name for the topic, but the moment you quote official OpenAI pricing, you need to map back to the current public naming surface. That is what keeps naming noise from turning into pricing confusion.
Why the official OpenAI price is not one number
The most common mistake on this topic is not misunderstanding one number. It is acting like there should be only one number in the first place.
OpenAI's public pricing for the current GPT Image route is split across two useful layers:
- The main pricing page gives the token billing model.
- The image generation guide gives per-image output pricing.
Both are real. They just answer different cost questions.
First layer: token billing on the main pricing page
As of April 17, 2026, OpenAI's API Pricing page lists these public token prices for gpt-image-1.5:
| Billing item | Current public price |
|---|---|
| Text input | $5.00 / 1M tokens |
| Cached text input | $1.25 / 1M tokens |
| Image input | $8.00 / 1M tokens |
| Cached image input | $2.00 / 1M tokens |
| Text output | $10.00 / 1M tokens |
| Image output | $32.00 / 1M tokens |
This is the right layer to use when you want to understand total request cost. If your workflow includes text prompts, image edits, image input, or repeated prompt reuse, the token table is the real billing baseline.
Second layer: per-image output pricing in the image guide
The OpenAI image generation guide gives a second table that is much easier to read when you want a rough "how much does one image cost?" answer:
| Size | Low | Medium | High |
|---|---|---|---|
1024 x 1024 | $0.009 | $0.034 | $0.133 |
1024 x 1536 | $0.013 | $0.05 | $0.2 |
1536 x 1024 | $0.013 | $0.05 | $0.2 |
The same guide also states that GPT Image Latest currently uses the same documented per-image output pricing as GPT Image 1.5.
That does not mean one output row is the whole bill. The per-image table is useful because it gives a quick cost estimate for the generated image itself. But text input and image input costs still sit on top of that. If someone compresses all of this into one "official GPT-Image-2 price," they are already removing the part of the pricing model that matters most once usage becomes real.
The safest way to read the official price today
If you want a simple rule that survives all the naming noise, use this order:
- If you need the official current billing truth, start with
gpt-image-1.5on the main OpenAI pricing page. - If you want the intuitive cost of one generated image, add the per-image output pricing from the image generation guide.
- If someone gives you only one unlabeled number, assume context is missing until you know whether it came from token billing, per-image output pricing, or a third-party route.
Why some gpt-image-2 pages are cheaper
Cheap gpt-image-2 pages do not automatically mean OpenAI has a second hidden official row. In most cases, they are selling something different:
- a relay route
- a compatible wrapper API
- a reseller route
- a domestic payment layer
- an internally simplified per-call pricing model
Any of those routes can be useful for the right buyer. If you only want to test a compatible endpoint quickly, or you care more about payment convenience than perfect alignment with OpenAI's own billing surface, a third-party route might be a reasonable choice.
But that choice belongs to a different question. Once you move from "what does OpenAI publicly price this at?" to "what is the cheapest way for me to get something close to this capability?", you are no longer comparing two rows inside the same table. You are comparing two contracts with different stability, support, logging, reliability, and pricing boundaries.
That is why the order matters:
- Establish the official OpenAI baseline first.
- Decide whether your goal is official billing accuracy or cheaper access.
- Only then compare third-party route pricing.
Reverse that order and every low-cost page starts to look like a secret official discount.
If you need a broader price comparison
If your actual job is to compare multiple image APIs rather than sort out the current official OpenAI row name, start with our AI Image Generation API Comparison 2026.
If you are trying to figure out the cheapest official OpenAI starting point, free-trial boundary, or low-cost API onboarding route, go to OpenAI API Key Free Trial 2026.
Use those pages for broader route selection. Start here only to lock the official pricing baseline and separate it from cheaper access routes.
FAQ
What is the relationship between GPT Image Latest and gpt-image-1.5?
In the public OpenAI image generation docs as of April 17, 2026, gpt-image-1.5 is the latest and most advanced current public GPT Image model. The same image guide says GPT Image Latest currently uses the same documented per-image output pricing. For pricing purposes, they land in the same public pricing layer today.
Why does official pricing split token billing and per-image output pricing?
Because they answer different cost questions. Token billing is closer to the real total request bill. Per-image output pricing is closer to the intuitive cost of generating one image at a given size and quality. You need both if you want a correct read on cost.
Do cheap gpt-image-2 quotes count as official OpenAI pricing?
No. They can be valid offers for a third-party route, but they are not the same thing as OpenAI's public official price table. Treat them as route-specific access pricing, not as a hidden official row.
If I only care about the cheapest image generation route, what should I do next?
After you lock the official baseline, compare broader options in AI Image Generation API Comparison 2026, or check the official low-cost OpenAI starting path in OpenAI API Key Free Trial 2026.
Bottom line
The shortest correct answer is this:
When you see gpt-image-2, do not read it as the current public OpenAI pricing row. The official price to trust today maps to gpt-image-1.5 / GPT Image Latest, not to a separate hidden gpt-image-2 row.
Once that boundary is clear, the rest of the price conversation becomes much easier. You can trust the official table for billing truth, then make a separate decision about whether a cheaper third-party route is worth the tradeoff.