OpenAI's Image AI Makes GIFs Now: Images 2.6, the Limits and the Catch

OpenAI's image model can make animated GIFs now. Images 2.6 came out on November 4, 2026, in ChatGPT, Codex and the API, and OpenAI's own pitch is "turning a product shot into an eye-catching loop."

What changed

OpenAI announced Images 2.6 on its blog and listed it in the API changelog on November 4, 2026. In ChatGPT and Codex it's available on all subscription tiers, per OpenAI. In the API there are two models, GPT-Image-2.6 Flare (fast iteration) and GPT-Image-2.6 Sunburst (detailed work). The new pieces:

• Animated GIFs. Set output_format: "gif" on the generations or edits endpoint and describe the subject, motion and loop in your prompt. To animate from your own still images, send them to the edits endpoint as references.
• Frame controls. frame_count takes 2 to 16 frames (default 9), and frame_duration_ms takes 20 to 2,000 ms per frame (default 200). OpenAI's example lantern GIF, made with Sunburst, uses 9 frames at 200 ms, so it loops every 1.8 seconds.
• SVG export. Set output_format: "svg" on the generations endpoint (with n set to 1) for icons and illustrations with editable paths, shapes and colors.
• Markup-guided edits. Draw on an overlay to show where a change should go. It works on the edits endpoint with supported 2.5 and 2.6 models.
• A complexity setting. low, medium or high (the default) controls how much the model plans before it renders. It has no separate fee.

OpenAI also says Images 2.6 generates images 50% faster than before, and that Flare can return an image in as little as six seconds at low quality. Those are OpenAI's numbers. I haven't timed them.

What it costs

The API pricing page lists both 2.6 models at the same per-token rates as the 2.5 models: $30 per million image output tokens, $8 per million image input tokens ($2 cached) and $5 per million text input tokens. That's a rate, not a price per GIF. A GIF's cost depends on how many tokens it uses, so run a few test GIFs before you quote anyone.

The catch

• Text. OpenAI's guide says the model "can still struggle with precise text placement and clarity." Don't trust it with prices or product names.
• Consistency. The same guide says it may struggle to keep recurring characters or brand elements consistent across generations.
• Speed on hard prompts. Complex prompts may take up to 2 minutes.
• GIF limits. GIF output doesn't support streaming, partial images, custom compression or the prompt rewriter.
• Access. You may need to complete API Organization Verification before using GPT Image models.
• Old models are leaving. This part isn't new: in June, OpenAI announced that gpt-image-1.5, gpt-image-1-mini and chatgpt-image-latest shut off on December 1, 2026. Its deprecations page names the 2.6 models as the replacement. If your app still calls one of them, move it before then.

Here's how you get paid

My opinion, no income guaranteed: many small shops already have product photos and send emails. Offer them a pack of looping product GIFs:

1. Their photos in. Send the shop's own product shots to the edits endpoint as reference images, with output_format: "gif".
2. Loops out. One short loop per product, sized for their email template.
3. Text by you. Add prices, names and any text yourself in an editor, on top of the finished loop, since the model can still fumble text.

Check every frame before it ships, and run your own cost test so your price covers the tokens.