The two new image models take first and second

The Artificial Analysis post is timestamped 2026-11-06T23:40:40+00:00, which is 07:40 in Beijing on 7 November. The post says OpenAI's GPT Image 2.6 takes first and second on both AA-Image-T2I v2.1 and AA-Image-Editing v2.1. It says GPT Image 2.6 Flare is almost twice as fast as GPT Image 2.5 Flare, with slightly better quality, and that GPT Image 2.6 Sunburst takes the top spot with a longer generation time.

The post dates the model release to 4 November, in ChatGPT, Codex, and the API. The API offers Flare and Sunburst, following the same names on 2.5. Both support text to image and image editing, and the post says both add animated GIF and SVG output. It also says the API adds a complexity setting that controls how much planning happens before rendering. The test used maximum quality for both, with complexity at the API default of high. That is the condition this organization wrote down, not a procedure to repeat.

The times come from the same post. Flare's median text-to-image time fell from about 55 seconds to about 30 seconds, and its editing time fell from about 58 seconds to about 32 seconds. Sunburst moves the other way: slower than GPT Image 2.5 Sunburst, at about 2 minutes 15 seconds per image, and in first place. The post says the largest text-to-image gains are in Reasoning, Text Rendering, Lighting, and Material, and that the strongest area is UIUX Design.

The price is the same as GPT Image 2.5: 30 dollars per 1 million image output tokens. The post converts that to about 0.21 dollars per 1024 by 1024 image at maximum quality, or about 211 dollars per thousand images. Availability is written as every tier of ChatGPT and Codex, plus the OpenAI API.

The post's words do not list Elo scores one by one. The same post attaches a ranking chart that prints scores. This column uses only the ranks, times, and prices written in the text. The post does not state a sample size or a vote count.