AI SDK OpenAI provider adds GPT Image 2.6 with animated GIF output (November 5, 2026)

What changed

On November 5, 2026 @ai-sdk/openai 4.0.109 added support for the GPT Image 2.6 models, gpt-image-2.6-flare and gpt-image-2.6-sunburst (the AI SDK 6 line got the same change in @ai-sdk/openai 3.0.137). New image provider options:

• complexity: 'low', 'medium' or 'high', how much detail goes into the image, for generation and editing. The API default is high.
• markup (editing only): a transparent PNG, as an HTTPS URL or base64 data URL, laid over the first input image to mark the areas to change. Works with gpt-image-2.5-* and gpt-image-2.6-*. It can't be combined with mask.
• outputFormat: 'gif' (generation only): an animated GIF. frameCount sets the number of frames (default 9; the AI SDK docs say 2 to 64, but OpenAI's image guide allows 2 to 16) and frameDurationMs sets 20 to 2000 ms per frame (default 200).
• safetyIdentifier: an ID for your end user, up to 128 characters, that helps OpenAI detect abuse.

Prerequisites

• An OpenAI API key in OPENAI_API_KEY
• ai 7 and @ai-sdk/openai 4.0.109 or later (4.0.111 was the newest on November 7)

Steps

1. Upgrade: pnpm add ai @ai-sdk/openai
2. Generate an animated GIF. This is the provider docs example:

const { image } = await generateImage({
  model: openai.image('gpt-image-2.6-flare'),
  prompt: 'A red panda waving hello, flat illustration',
  providerOptions: {
    openai: {
      outputFormat: 'gif',
      frameCount: 9,
      frameDurationMs: 200,
    } satisfies OpenAIImageModelGenerationOptions,
  },
});