OpenAI Images 2.6: 6-Second Generation, 50% Speed Boost with Flare and Sunburst Tiers

On November 4, OpenAI released Images 2.6, a major update to its image generation models that delivers a 50% speed improvement over the previous version.

The release introduces two distinct model tiers:

Flare (gpt-image-2.6-flare): Optimized for speed. At the lowest quality setting, it can generate an image in as little as 6 seconds . Available to all ChatGPT and Codex subscribers across desktop, mobile, and web.

Sunburst (gpt-image-2.6-sunburst): Optimized for quality and stability during iterative editing. OpenAI states it produces finer detail at medium/high quality settings and handles multiple revision rounds more reliably, making it suitable for posters and product images. Exact latency numbers for Sunburst were not disclosed.

OpenAI reports that across ChatGPT, Codex, and the API, the combined system now generates 4 billion images per week , averaging over 500 million images per day . The 6-second figure represents Flare's peak speed at reduced quality; not every image completes that quickly. The weekly volume reflects total output across all three channels, not incremental growth from this release alone.

Pricing (Flare)

Flare pricing is published per million tokens:

Image output: $30

Image input: $8

Cached image input: $2

Text input: $5

Cached text input: $1.25

Sunburst pricing is not included in the published table.

API Capabilities and Parameters

The API supports several advanced features:

Looped animations : Only available at generation time by requesting GIF format. Frame count configurable from 2 to 64 (default 9). Frame duration 20–2000 ms (default 200 ms), rounded to the nearest 10 ms. The edit endpoint cannot produce animations.

Vector images : Only generated at creation time, one per request, returned as static images without scripts.

Partial editing via mask : Attach a transparent image of the same dimensions to indicate the region to modify.

Complexity parameter : Accepts low, medium, high (default high) with no extra charge. The documentation recommends low for most edits. Image clarity remains controlled by the quality parameter.

Sources

OpenAI blog announcement: https://openai.com/index/introducing-images-2.6/

Flare model documentation: https://developers.openai.com/api/docs/models/gpt-image-2.6-flare

Image generation guide: https://developers.openai.com/api/docs/guides/image-generation