GPT Image 2.6 is in Hermoso, and Flare drafts in about 11 seconds

OpenAI released Images 2.6 on November 4. A day later it's in Hermoso as two models: GPT Image 2.6, now our default for image ads, and GPT Image 2.6 Flare, a fast version for when you want ten drafts before lunch.

If you've been looking for "the ChatGPT image model" in Hermoso, this is it. GPT Image is the name OpenAI gives the model family behind image generation in ChatGPT.

Why it matters for ads

An ad image lives or dies on its words. The headline, the offer, the price on the sticker, the small print on the label. GPT Image models have been the best at putting real, readable text into a picture, and 2.6 is the sharpest yet on text and fine detail. Fewer renders get thrown away because a letter came out wrong.

It also keeps your product real. Attach up to 16 reference images (your packshot, your logo, a photo of the shop) and describe the scene around them.

Flare is the fast one

In our own test, a GPT Image 2.6 Flare image came back in about 11 seconds. The full GPT Image 2.6 took about 50. So the workflow that makes sense: draft a batch of ideas on Flare, pick the two or three that work, then render the finals on 2.6.

Both cost the same as the GPT Image 2.5 models they replace, and both make all 10 aspect ratios, from 9:16 for Reels and TikTok to 4:5 for the feed and 21:9 for a banner.

How to use it in the web app

• Just ask. In the Studio, ask for a static ad. Image ads now go to GPT Image 2.6 unless you pick something else.
• Pick it. Choose GPT Image 2.6 or GPT Image 2.6 Flare from the model menu, or open it in the Models tab and type a prompt.
• Edit. Upload a photo and ask for the change you want. GPT Image 2.6 edits with up to 16 reference images.

How to use it from ChatGPT, Claude or any agent

Connect Hermoso to your assistant over MCP and ask in plain English: "Make three 4:5 static ads for our cold brew with GPT Image 2.6, headline 'Up all night'." Agents that call tools directly use generate_image with model: "gpt-image-2.6" or "gpt-image-2.6-flare". Leave the model out of render_ad and an image ad renders on 2.6 anyway.

Building your own tool? The same models are on the Hermoso API, with the same model names.

What happened to GPT Image 2.5?

It's the previous version now. Its pages point to 2.6, and new image ads start on 2.6. Product photos and YouTube thumbnails still default to Nano Banana 2.1, which is still the stronger pick for a true-to-life product on a set. You can always name the model you want.

See every image model, best first, on the image models page.

Questions people ask

Is GPT Image 2.6 the same model as ChatGPT Images 2.6?

GPT Image is the name OpenAI gives the model family behind image generation in ChatGPT, and 2.6 is the version OpenAI released on November 4, 2026. In Hermoso it is called GPT Image 2.6, with a fast version called GPT Image 2.6 Flare.

What is the difference between GPT Image 2.6 and GPT Image 2.6 Flare?

Flare is the fast version. In our test a Flare image took about 11 seconds and a GPT Image 2.6 image about 50. Use Flare for drafts and 2.6 for the final, sharpest render.

Does GPT Image 2.6 cost more than GPT Image 2.5?

No. Both 2.6 models cost the same as the 2.5 models they replace. The exact price of every render is shown in the app before it runs.

Can I use GPT Image 2.6 from ChatGPT, Claude or my own code?

Yes. Connect Hermoso over MCP and ask for an image in plain English, or call generate_image with the model gpt-image-2.6 or gpt-image-2.6-flare. The same model names work on the Hermoso API.