GPT Image 2.6: Flare vs Sunburst, GIFs, Speed & Price

You can use GPT Image 2.6 in Sumus today — just ask your AI teammate to make an image or GIF. GPT Image 2.6 is OpenAI's new image model, and the headline is simple: ChatGPT can now make GIFs. OpenAI launched Images 2.6 on November 4, 2026, for ChatGPT, Codex and the API. The update is 50% faster than the last version, turns a prompt or a photo into an animated GIF, and lets you save reference images of yourself, your pets or your characters.

Sumus also puts GPT, Claude, Gemini and Grok in one chat, so one model can write the prompt and another can check the image.

For developers, it ships as two models, GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst. Below: what changed, Flare vs Sunburst, API features, price, safety and how to try it. Every fact comes from OpenAI's launch post, the system card and OpenAI's own X thread.

GPT Image 2.6 at a glance

What's new in GPT Image 2.6

• GIFs. Turn a prompt or a photo into an animated GIF, change the playback speed, edit single frames and share it on repeat. Before this, ChatGPT could not make a GIF in one step.
• Reference images. "Save once. Create again and again." Keep photos of yourself, pets or characters ready to reuse without uploading them again.
• Speed. OpenAI says the new version makes images 50% faster than before.
• Better edits. The system card says both models improve edit consistency, detail preservation and how they follow sketches and image notes.
• Layouts. Flare improves layout quality for posters, product shots and other pro visuals.

OpenAI also shared a usage number: people now make 4 billion images a week, which is more than half a billion a day. The Verge notes that was 3 billion a week in September.

GPT Image 2.6 Flare vs Sunburst

Both models first appeared as GPT Image 2.5 in September; 2.6 is the upgrade of both. A simple rule: start on Flare, and move to Sunburst only when your edits start to drift.

GPT Image 2.6 GIFs: how they work

Ask for a GIF in plain words, like OpenAI's demo "can you make a gif of a cat dancing?", or upload a photo or even a pen sketch. The model draws a set of frames that keep the same character, then plays them as a loop. OpenAI's own example prompts ask for "a single animation sprite sheet" of 6 frames in a 3-by-2 grid, which shows how the model builds a GIF under the hood. Kingy.ai, reading the API docs, reports 2 to 64 frames per GIF. For a step-by-step guide, read our ChatGPT GIF guide.

GPT Image 2.6 in the API

• Image markup: draw on an image to show exactly where the change goes. It was in ChatGPT; now it is in the API.
• SVG export: get vector files ready for websites, apps and print.
• Complexity: a new setting for how much planning the model does. OpenAI suggests low for quick edits and medium or high for busy scenes.
• Quality: still controls detail and fidelity, separate from complexity.

Early partners include OpenArt, Figma Weave, Picsart, fal, Higgsfield and Town, a startup that used it to animate Monty, the typing character in its work assistant.

GPT Image 2.6 price

OpenAI points to its API pricing page for GPT Image 2.6. Image models there are billed per token: GPT Image 2.5 Flare and Sunburst list $8 per 1M image input tokens and $30 per 1M image output tokens, plus $5 per 1M text input tokens. Kingy.ai reports the same rates for the 2.6 versions. In ChatGPT, Images 2.6 is included on every plan, Free included.

GPT Image 2.6 safety

The system card says the safety stack is the same as Images 2.5, with extra checks for GIFs, since motion can carry meaning a still frame does not. A safety model reviews the full frame grid and caption. On 310 rule-breaking test GIFs it caught 77.66%. Overall, both new models scored close to their 2.5 versions.

Try GPT Image 2.6 next to other image models in Sumus

Yes, you can compare its output with other models without juggling tabs. Sumus is one chat where AI teammates on Claude, GPT, Gemini, Grok and DeepSeek work together. Ask one teammate to write the prompt, another to make the image, and a third to check it, and your team sees every result in the same thread.