Skip to content

OpenAI shuts down its older image models on 1 December. Will your image automations break?

By Bilal Haider, Founder, Theepa Studio · · 5 min read

Short answer

OpenAI removes gpt-image-1-mini, gpt-image-1.5 and chatgpt-image-latest from its API on 1 December 2026, and gpt-image-2 on 10 February 2027. Its recommended replacements are gpt-image-2.6-sunburst and gpt-image-2.6-flare. OpenAI does not say old calls will be rerouted, so find every image step that names these models and test a 2.6 model before 1 December.

Which OpenAI image models are shutting down?

OpenAI's API deprecations page, read on 12 November 2026, lists three image models for removal from the API on 1 December 2026: gpt-image-1-mini, gpt-image-1.5 (including the gpt-image-1.5-2025-12-16 snapshot) and chatgpt-image-latest. OpenAI says it notified developers on 2 June 2026. On 10 November 2026 it added a second notice: gpt-image-2 and its gpt-image-2-2026-04-21 snapshot will be removed on 10 February 2027.

ModelRemoved from the APIOpenAI's recommended replacement
gpt-image-123 October 2026 (already gone)gpt-image-2.5-sunburst or gpt-image-2.5-flare
gpt-image-1-mini1 December 2026gpt-image-2.6-sunburst or gpt-image-2.6-flare
gpt-image-1.51 December 2026gpt-image-2.6-sunburst or gpt-image-2.6-flare
chatgpt-image-latest1 December 2026gpt-image-2.6-sunburst or gpt-image-2.6-flare
gpt-image-210 February 2027gpt-image-2.6-sunburst or gpt-image-2.6-flare

The notice covers the OpenAI API: the models that automations, plugins and scripts call with an API key. It does not describe changes to the ChatGPT app.

Will my image automations stop working on 1 December?

Plan as if they will. When Google retired Gemini 2.5, it said old calls would be rerouted to newer models. OpenAI's page says no such thing. It says these models will be removed. Any step that names one of them, in Zapier, Make or n8n, in a website plugin or in your own code, should be expected to fail from that date. In a small business, image steps tend to hide in places like these:

  • a workflow that turns each new blog post or product into a social graphic
  • a product photo step that removes or swaps backgrounds
  • a design or marketing tool that asked for your own OpenAI API key
  • a script someone wrote last year that nobody has opened since

If you use gpt-image-2 today, nothing changes on 1 December, but you have a second deadline in February. Our advice: skip a step and move straight to a 2.6 model, so you test and approve one change instead of two.

Should you choose GPT Image 2.6 Sunburst or Flare?

OpenAI positions Sunburst for detailed compositions and image editing, and Flare for fast, interactive iteration. Both model pages also mention a complexity setting, markup-guided edits, SVG output and animated GIF output. Two details matter before you switch.

Endpoints. The 2.6 model pages list the Images API generation and edit endpoints as supported, and the Responses, Chat Completions and Batch endpoints as not supported. If your workflow makes images through the Responses API or in batch jobs, check OpenAI's image generation guide for the supported route first.

Price. Both 2.6 models have the same list price. Here are OpenAI's standard prices per 1 million image tokens, read on 12 November 2026 (text input is priced separately):

ModelImage inputImage outputCached image input
gpt-image-1-mini$2.50$8.00$0.25
gpt-image-1.5 and chatgpt-image-latest$8.00$32.00$2.00
gpt-image-2$8.00$30.00$2.00
gpt-image-2.6-sunburst or gpt-image-2.6-flare$8.00$30.00$2.00

If you were on gpt-image-1.5, the price per token barely moves. If you picked gpt-image-1-mini to keep costs down, the output price per token on a 2.6 model is almost four times higher. OpenAI's pricing page gives no price per image, because tokens per image depend on size and settings, so run your real prompts through the calculator in OpenAI's image generation guide before you commit. One more detail: gpt-image-2 has a batch rate at half its standard price, and the 2.6 models are not in the batch price table.

How do I keep AI images on brand after the switch?

A new model draws differently. Colours drift, text renders differently, and products and people can look new. Our advice:

  1. 1Never ask the model to draw your logo. Place the real logo file on top as a step in the workflow.
  2. 2Put your rules in the prompt: brand colours, type and the do and don't list from your brand guidelines.
  3. 3Keep 10 approved images from the old model as a reference set, and compare new results against them.
  4. 4Keep a person approving images before anything is published.

If your brand has no clear rules for a model to follow, our brand identity work produces guidelines written for people and for tools. There is more on this in is your brand system ready for AI design agents?

Worked example: a café group's weekly social graphics

This is a hypothetical scenario, not a client.

A café group with four branches runs two image automations. An n8n workflow uses gpt-image-1-mini to make three menu graphics a week for Instagram. A Make scenario uses gpt-image-1.5 to clean up backgrounds on new product photos.

  1. 1This week: inventory. The marketing lead opens both automations and notes the model, image size and quality setting each step sends. She also searches the team's scripts for "gpt-image".
  2. 2Copy, don't edit. She duplicates both. The menu graphics copy moves to gpt-image-2.6-flare, where speed matters. The product photo copy moves to gpt-image-2.6-sunburst, where detail matters.
  3. 3Test with real inputs. She runs last month's 12 menu prompts and 10 product photos through the old and new versions and checks colours and logo placement against the brand guide.
  4. 4Check the cost. She compares the test spend on her OpenAI usage page. The menu graphics move from the mini price to the full price, so she decides whether three a week still earn their cost, or whether one designed template with changing text would do the job.
  5. 5Switch by 24 November. The new versions go live a week before the deadline. The old ones stay paused, not deleted, until 1 December has passed.

What should I do this week?

  1. 1Search code, automations and plugin settings for "gpt-image" and "chatgpt-image-latest".
  2. 2List every tool that holds your OpenAI API key.
  3. 3Mark each step with its deadline: 1 December for gpt-image-1-mini, gpt-image-1.5 and chatgpt-image-latest, 10 February for gpt-image-2.
  4. 4Test gpt-image-2.6-flare and gpt-image-2.6-sunburst with real prompts and the brand checks above.
  5. 5Switch before the last week of November.

If your list runs past a few steps, our AI workflows and automation service maps every automation, moves each one to the right model and leaves your team a setup it can run. For text models, the next OpenAI deadline is 14 January 2027: see our guide to the GPT-4o shutdown.

Sources

Bilal Haider

Founder, Theepa Studio

Bilal founded Theepa Studio in 2016. The studio has delivered more than 10,000 brand identities for companies in the US, UK and the GCC.

Questions people also ask

When does OpenAI shut down gpt-image-1.5 and gpt-image-1-mini?

On 1 December 2026. OpenAI's deprecations page lists that removal date for gpt-image-1-mini, gpt-image-1.5 (including the gpt-image-1.5-2025-12-16 snapshot) and chatgpt-image-latest, and recommends gpt-image-2.6-sunburst or gpt-image-2.6-flare instead.

Is gpt-image-2 affected?

Yes, later. On 10 November 2026 OpenAI announced that gpt-image-2 and its gpt-image-2-2026-04-21 snapshot will be removed from the API on 10 February 2027, with the same 2.6 replacements.

Will calls to the old image models be rerouted automatically?

OpenAI's deprecations page does not say so. It says the models will be removed from the API, so plan for any step that names them to stop working on the shutdown date.

Is the 2.6 replacement more expensive?

It depends on where you start. On OpenAI's pricing page, read on 12 November 2026, the 2.6 models cost $30 per 1 million image output tokens, against $32 for gpt-image-1.5 and $8 for gpt-image-1-mini. Teams moving from the mini model should test real prompts and check the cost.

Keep reading

Tell us what you're building.

A 30-minute call. We look at your brand and site together, and you leave with a plan and a fixed price, whether or not you hire us.