ABAB News Logo
Back to Flash News

OpenAI Releases Images 2.6: Supports GIF and SVG Export, with 50% Faster Generation

OpenAI released the image model Images 2.6 on November 4. It is now available across all subscription tiers in ChatGPT and Codex, covering desktop, mobile, and web. Users can generate GIF animations, and overall image generation is 50% faster than the previous version.

The official announcement said Images 2.6 supports bringing images to life dynamically, such as making brand mascots or characters move, or turning product images into looping animations. It also said editing can better preserve consistency in characters, products, and brand details. OpenAI said users worldwide generate 4 billion images per week on ChatGPT, Codex, and the API, averaging more than 500 million per day.

On the API side, developers can choose between two versions: GPT-Image-2.6 Flare is speed-oriented, and at low quality it can generate an image in about 6 seconds at the fastest, while also improving layouts for posters and product images; GPT-Image-2.6 Sunburst is oriented toward image quality and editing consistency, with more detail at medium or high quality, and characters and products are less likely to go off-model during multiple rounds of edits. These speed and quality descriptions all come from OpenAI.

Developers can also use image annotation via the API (precisely modifying selected areas by circling them), the complexity parameter (which controls how strongly the model plans the composition), and SVG export. OpenAI said in the announcement that SVG export allows designs to be used directly for websites, apps, and printed materials. The announcement did not specify whether the ChatGPT interface directly supports SVG export. Figma Weave and Picsart have already integrated it in advance.

Images 2.6 is available in all tiers of ChatGPT and Codex, and the announcement did not mention any additional fee. The API pricing page shows Flare and Sunburst are priced the same: image output costs $30 per 1 million tokens, image input costs $8 per 1 million tokens, and text input costs $5 per 1 million tokens, unchanged from the previous generation 2.5. 2.6 is not yet listed in the batch processing price table.

Original source:

  1. Title: Introducing Images 2.6 Publisher: OpenAI Publication date: 2026-11-04 URL: https://openai.com/index/introducing-images-2-6/ Source nature: official announcement Verification status: full text key passages read this time, verification date 2026-11-04 Supported content: Images 2.6 launched on ChatGPT/Codex/API, all tiers, desktop/mobile/web; 50% faster; GIF feature and use cases; Flare/Sunburst distinction and 6-second statement; SVG, annotation, complexity parameter; 4 billion images per week; pricing page link

ABAB AI Insight

The Images series has progressively strengthened editing consistency and speed from 2.0 to 2.5. When 2.5 was launched in September 2026, it also offered the dual Flare (speed) and Sunburst (precision) versions, and claimed latency was reduced by up to 50% versus 2.0. 2.6 continues this dual-track design while formally adding GIF and SVG to API capabilities, indicating that OpenAI is pushing its image model from one-time image generation toward an iterative, deliverable production tool.

In terms of capital and product strategy, weekly usage of 4 billion images has already made images a high-frequency consumption scenario for OpenAI. API pricing is based on tokens and remains unchanged from 2.5, meaning that faster generation and new formats do not come with a price increase; on the ChatGPT side, all tiers are covered, lowering the barrier to use. Early integration by Figma Weave and Picsart shows that design and creation tools are embedding the model into existing workflows rather than using it only as a standalone chat feature.

Compared with Google Nano Banana 2.1, both products have the same image output price of $30 per 1 million tokens. OpenAI additionally offers SVG export and complexity control, more directly targeting design scenarios that require editable vector output and precise local modifications; Google, meanwhile, provides a more explicit per-image estimate on its pricing page (about $0.034 for 1K). Both are competing for the entry point of image generation that can be used directly in products.

For OpenAI, GIF and SVG reduce the conversion cost from AI drafts to deliverable assets: users or developers no longer need to trace bitmap images into vectors again, or separately use other tools to create simple animations. The key metrics to watch are the share of SVG and GIF output in subsequent API calls, and actual usage growth after design-tool integration—if these formats’ share rises, it would indicate that the model is shifting from presentation-oriented generation toward production assets.

ABAB News · Cognition Law

  1. For image models, moving from bitmap output to SVG and GIF support is a step from making results that are just for viewing to results that can be used directly.

  2. Separating speed and precision while keeping prices unchanged means using faster iteration to win more calls under high-frequency usage, rather than simply raising prices.

  3. Early integration by design tools shows that the value of image capabilities is shifting from the chat interface to production workflows.