Seedance 2.5 is live on EvoLinkTry Seedance 2.5
Abstract AI model release verification system with a sealed luminous core and status checkpoints
Product Launch

GPT Image 2.5 Release Date and Current Status

EvoLink Team
EvoLink Team
Product Team
September 4, 2026
14 min read
As of September 4, 2026, OpenAI has not announced GPT Image 2.5. No release date, public model ID, price, or API entry exists. The newest OpenAI image model is gpt-image-2, released April 21, 2026. Community signals suggest a next image release is being prepared; nothing official confirms it.

GPT Image 2.5 status at a glance

QuestionStatus on September 4, 2026Evidence
Is GPT Image 2.5 released?NoOpenAI's model catalog, pricing page, API changelog, and SDK model types list no image model newer than gpt-image-2
Is there a release date?NoNo OpenAI announcement, teaser, or changelog entry
Is there a model ID or price?NoAny gpt-image-2.5 string or price you see today is third-party guesswork
Is OpenAI preparing an image upgrade?Unconfirmed; community signals onlyCommunity-reported imagegen25 announcement string (September 3) and anonymous Arena checkpoints (first reported August 9)
Is mona-lisa-1 GPT Image 2.5?UnknownNo official mapping; see the Mona Lisa 1 release status and Arena test
Is GPT Image 2 released?YesOfficial model page and pricing since April 21, 2026
Can I call GPT Image 2.5 through EvoLink?NoTrack the GPT Image 2.5 API status page for route verification

Timeline: from GPT Image 1.5 to the imagegen25 string

DateEventWhat it supportsWhat it does not support
December 16, 2025OpenAI releases GPT Image 1.5 in ChatGPT and the APIOpenAI has used a .5 name for an image mid-cycle updateThat the next name will also be .5
April 21, 2026OpenAI releases GPT Image 2, snapshot gpt-image-2-2026-04-21The current production baselineThe next release date
August 9, 2026Community first reports the anonymous label mona-lisa-1 in Arena Battle ModeAn unreleased image checkpoint was being testedProvider, final name, price, API
August 11, 2026EvoLink generates an output from mona-lisa-1 in Arena Battle ModeA callable anonymous checkpoint existed in that sessionAny link to a product name
August 20, 2026Community posts describe a second label, luna-lisa-alpha, with less visible noiseAnother test variant may existAny relationship to Mona, OpenAI, or a 2.5 name
August 20, 2026OpenAI changelog adds transparent-background preview for gpt-image-2OpenAI is still shipping features on GPT Image 2A successor date
September 3, 2026Community members report a feature-gated announcement modal in OpenAI's web client code with the internal identifier imagegen25Announcement copy for an image upgrade was reportedA product name, API ID, price, or date
September 4, 2026OpenAI catalog, pricing, changelog, and SDK types still show no image model newer than gpt-image-2GPT Image 2.5 remains unreleasedThat it will not ship soon

The release cadence is suggestive: about eight months from GPT Image 1 to 1.5, four months from 1.5 to 2, and four and a half months since GPT Image 2. Cadence raises the prior for a near-term update. It is not evidence of a date.

What the imagegen25 string can and cannot prove

On September 3, 2026, community members reported finding a disabled announcement modal in OpenAI's web client code. The reported component is imagegen-25-announcement-modal, with the internal identifier imagegen25, and the copy reads "Image creation got a major upgrade" with "Higher-quality results, faster generation, and smarter creative tools". The modal is reported as feature-gated, meaning no user would see it in production.

What this supports, if the report is accurate: OpenAI prepared announcement copy for an image upgrade, and the internal label uses "25".

What it does not support: a product called GPT Image 2.5, an API model ID, a price, or a date. Internal labels can be version numbers of a ChatGPT feature rather than a new API model. The original screenshots circulate on social platforms and could not be independently re-verified for this update, and the string does not correspond to any model ID in OpenAI's API catalog or SDK model types.

Why Mona Lisa 1 and Luna Lisa Alpha are not "GPT Image 2.5"

mona-lisa-1 is an anonymous label EvoLink observed directly in Arena Battle Mode on August 11, 2026. Arena's policy lets providers test unreleased models anonymously and remove them after collecting votes, so an anonymous label is a real pre-release signal but not a release promise, and not proof of who built the model.

Community reports say some outputs carried OpenAI provenance signals. OpenAI's own documentation states that C2PA and SynthID signals indicate an association with OpenAI tools, and that coverage varies by product, model, export path, and file type. A provenance signal can support "made with an OpenAI tool"; it cannot identify which model, and it cannot supply a commercial name.

luna-lisa-alpha and later labels are community reports only. Testers describe less noise and a more natural look, and they also describe the difference as incremental. Without a shared prompt set and original files, nobody outside the tests can separate a model-level improvement from favorable examples. The full first-hand record, including what EvoLink saw and what remains unknown, lives on the Mona Lisa 1 release status page.

Will it be called GPT Image 2.5 or GPT Image 3?

Both names circulate. Both names circulate in community coverage; neither has an official basis. The .5 theory rests on the GPT Image 1.5 precedent; the "3" theory rests on the size of the reported quality change. Some posts guessed at a launch alongside GPT-6 Astra; Astra shipped on September 3 with no image-model entry. Others point to OpenAI's DevDay; that is a guess with no OpenAI source behind it.
EvoLink uses "GPT Image 2.5" on this page because it is the name people search for. If OpenAI announces a different name, this page and the GPT Image 2.5 API status page will keep their URLs and state the mapping in the first paragraph.

ChatGPT Image 2 is released: the current baseline

Many people still search for "ChatGPT Image 2" because that is how they meet OpenAI's image generation inside ChatGPT. The official API model name is gpt-image-2, documented since April 21, 2026, and it is the baseline any successor has to beat.
What people searchWhat OpenAI calls itStatus
ChatGPT Image 2, ChatGPT Images 2.0gpt-image-2Released April 21, 2026
Older ChatGPT image generationgpt-image-1.5Listed by OpenAI as the previous image model; chatgpt-image-latest is a catalog alias, not a separate release
GPT Image 2.5, ChatGPT Image 2.5No official nameNot announced
What gpt-image-2 brought over the previous generation, from OpenAI's documentation: generation and editing through the Images API, low/medium/high quality tiers, PNG/JPEG/WebP output, any size up to 3840 px on the long edge with 1K/2K/4K presets and aspect ratios up to 3:1 (GPT Image 1.5 offered three fixed sizes), and, since August 20, 2026, a transparent-background preview.
OpenAI's published token pricing for gpt-image-2, shown here as the current baseline and not as a GPT Image 2.5 price:
Cost typePrice
Text input$5.00 per 1M tokens
Cached text input$1.25 per 1M tokens
Image input$8.00 per 1M tokens
Cached image input$2.00 per 1M tokens
Image output$30.00 per 1M tokens (GPT Image 1.5 was $32.00)
Third-party price tables commonly put a 1024×1024 output at roughly $0.006, $0.053, and $0.211 for low, medium, and high quality, before prompt text and input images; the exact figure depends on output tokens, and OpenAI points to its cost calculator rather than a fixed per-image price. For EvoLink route pricing and quality tiers, see the GPT Image 2 product page.

What a GPT Image 2.5 release would need to prove

OpenAI documents four limits for gpt-image-2: complex prompts can take up to about two minutes, text placement and legibility can still fail, characters and brand elements can drift across generations, and precise element placement in layout-sensitive scenes is limited. Community complaints add high-frequency noise, a plastic look on skin, and CJK text errors. The community-reported modal copy mentions higher quality and faster generation, which maps onto exactly these gaps.

A release is therefore worth adopting only if controlled tests show gains on:

  1. Product shots and packaging text: brand text accuracy, material fidelity, usable rate.
  2. Posters and layout-sensitive designs: text hierarchy and element placement.
  3. English and CJK in-image text: character accuracy and font consistency.
  4. Photoreal people: skin, hands, hard and low light, noise.
  5. Consistency across a set: identity and brand elements over many generations.
  6. Multi-turn edits and latency: drift after 3, 5, and 10 edits; p50 and p95 latency; timeout rate.

Judge on first-pass rate, cost per accepted image, and latency distribution, not on selected samples. EvoLink will publish this comparison only after a verified route exists.

Release does not mean API availability

Three channels move at different speeds, and a headline about one of them says nothing about the others. Model ID, pricing status, and EvoLink route verification are tracked on the GPT Image 2.5 API status page; this section only explains why the channels differ.
ChannelWhat "released" means thereStatus
ChatGPT productA new image experience appears in the app, possibly under a feature name rather than a model nameNot announced
OpenAI APIA model ID, price, and documentation appear in the catalogNot announced
EvoLink routeVerified model ID, approved pricing, one successful generation and one edit through the production route, and matching billingNot available
GPT Image 1.5 shipped to ChatGPT and the API together, and GPT Image 2's API snapshot carries its April 21 launch date. That is a pattern, not a commitment. If OpenAI ships a ChatGPT-only upgrade first, this page will say so, and the launch alert will stay in watch state until EvoLink's five verification gates pass.

Should you wait for GPT Image 2.5 or ship on GPT Image 2 now?

Your situationRecommendation
You need image generation or editing in productionShip on GPT Image 2. It is documented, priced, and runs on EvoLink today through the same endpoint a successor would use
You are planning an evaluationFreeze 30 to 100 representative prompts, source images, and acceptance rules now, with GPT Image 2 outputs as the baseline
You run a multi-model productKeep the model string in configuration and define the fallback route, so a verified successor is a one-line change
You only saw social examplesTreat them as test ideas, not purchasing evidence; no shared prompt set or original files are public
You saw a site selling "GPT Image 2.5 API access"No such API identifier exists; whatever it sells is a different model

Signals that will change this page

  • An OpenAI announcement, model page, pricing entry, SDK type, or changelog entry naming a new image model.
  • OpenAI clarifying that imagegen25 refers to a ChatGPT feature rather than an API model.
  • Arena or OpenAI disclosing the provider or identity of mona-lisa-1 or luna-lisa-alpha.
  • A verified EvoLink route with model ID, approved pricing, and a successful request.

EvoLink checks OpenAI's model catalog, changelog, pricing page, and SDK image model types on every update to this page. The last check date is in the update log below.

Where to go from here

Your goalStart here
Track model ID, pricing status, and EvoLink route verificationGPT Image 2.5 API status
Read the first-hand Arena test of the anonymous checkpointMona Lisa 1 release status
Use the current OpenAI image route todayGPT Image 2 on EvoLink
Compare GPT Image 2, 1.5, and 1GPT Image family
Integration guide with code examplesGPT Image 2 developer guide
Decide between GPT Image 2 and 1.5GPT Image 2 vs GPT Image 1.5
Browse prompt examplesGPT Image 2 prompts
Understand the "ChatGPT Image 2" vs gpt-image-2 namingChatGPT Image 2 release status

FAQ

Is GPT Image 2.5 released?

No. As of September 4, 2026, OpenAI's model catalog, pricing page, API changelog, and SDK model types list no image model newer than gpt-image-2. No model named GPT Image 2.5 has been announced.

When is the GPT Image 2.5 release date?

No date has been announced. Release cadence and community signals suggest an update is being prepared, but OpenAI has published nothing about timing.

What is imagegen25?

A community-reported internal identifier from a feature-gated announcement modal found in OpenAI's web client code on September 3, 2026. It indicates announcement copy for an image upgrade was prepared. It does not confirm a product name, an API model, a price, or a date.

Is Mona Lisa 1 the same as GPT Image 2.5?

That has not been confirmed. mona-lisa-1 is an anonymous Arena label that EvoLink observed directly. Arena allows anonymous testing of unreleased models, so the label proves neither the provider nor the final product name.

Will the next model be called GPT Image 2.5 or GPT Image 3?

Unknown. Both names circulate in the community. OpenAI used a .5 name for GPT Image 1.5, but naming history does not decide the next name.

Is ChatGPT Image 2.5 the same as GPT Image 2.5?

Yes, as search terms. "ChatGPT Image 2.5" and "ChatGPT Images 2.5" are how people describe a possible next image upgrade inside ChatGPT; "GPT Image 2.5" is the API-style name. Neither is an official OpenAI name today.

Is ChatGPT Image 2 released?

Yes. OpenAI released it on April 21, 2026, and the official API model name is gpt-image-2. "ChatGPT Image 2" is a search term, not an API identifier.

Is there a GPT Image 2.5 API or price?

No. Track the GPT Image 2.5 model ID and pricing status on the API status page.
Not yet. Join the launch alert and use gpt-image-2 today.

Should I wait for GPT Image 2.5?

For production work, no. Ship on GPT Image 2 with the model string in configuration and a frozen evaluation baseline. That is the fastest way to adopt a successor and the safest if it never ships under this name.

Sources and update log

Update log: September 4, 2026, first published as the GPT Image 2.5 release-status page with the evidence timeline, the imagegen25 signal, and the wait-or-ship decision.

Ready to Reduce Your AI Costs by 89%?

Start using EvoLink today and experience the power of intelligent API routing.