Up from more than 3 billion a week in 8 weeks
Thibault Sottiaux, who leads the programming assistant "Codex" at OpenAI, posted on X on Nov. 5 (Nov. 6 Japan time) that images created in ChatGPT and the API had topped 4 billion a week. Eight weeks earlier, the figure was more than 3 billion a week, he said. He quoted OpenAI's post about "ChatGPT Images 2.6," the new version of its image generation released the previous day.
Wild stat. We are now generate more than 4B images per week across ChatGPT and the API, up from the more than 3B per week only 8 weeks ago.
Calculated from 3 billion and 4 billion, the increase over 8 weeks is about 33%. Both are lower bounds stated as "more than," so the actual growth rate may differ from this calculation.
All of the figures come from the company. The split between ChatGPT and the API, the number of images by model and how a week is counted are not given in the materials that could be checked. No material in which a third party verified the numbers has been found either.
More than 500 million a day is a figure from media reports
The 4 billion-a-week figure had appeared in media reports before Sottiaux's post. The Next Web reported at the launch of Images 2.6 on Nov. 4 (Nov. 5 Japan time), citing OpenAI, that ChatGPT and the API produced more than 4 billion images a week, up from 3 billion in September. The Decoder also wrote that the figure was 1 billion more than in September.
Sottiaux's post does not give a daily figure. FoneArena and BigGo Finance reported, citing OpenAI, that the daily average was more than 500 million. Dividing 4 billion a week by 7 gives about 570 million, which is consistent with that figure. OpenAI's announcement could not be opened, so the company's wording could not be confirmed.
Scope differs between sources
Sources describe what the figure covers differently. Sottiaux's post and The Next Web say ChatGPT and the API. BigGo Finance and ChainCatcher add Codex, and the September reports referred to ChatGPT Images and the GPT-Image models in the API. Images 2.6 can also be used in Codex, but whether images made in Codex are included in the 4 billion has not been confirmed.
There are also unverifiable points about the comparison period. More than 3 billion a week was the figure reported when ChatGPT Images 2.5 launched on Sept. 8, close in time to the 8 weeks ago Sottiaux referred to. He has not said whether he was referring to this figure or whether the 2 figures were tallied using the same definition. How a GIF is counted and whether edited images are included have not been confirmed either.
No reason for the increase has been given. The 4 billion-a-week figure had already been stated on the day Images 2.6 launched, and the 8 weeks largely correspond to the period when Images 2.5 was available. Whether use after the 2.6 launch is included in the figure has not been confirmed.
Nov. 3 report said 1.5 billion a week
Big Technology, a U.S. newsletter, wrote on Nov. 3, as part of OpenAI's explanation of ChatGPT's growth, that users create 1.5 billion images a week. It came from the same material as the article on the scale of ChatGPT's search, and the figure is less than half of 4 billion a week. The speaker was not identified, and it was not stated whether the API was included. No material has been found to explain the gap between the 2, such as whether it is a figure for ChatGPT alone or an older figure.
- OpenAI says more than 700 million images were made in the first week after the March 2025 launchOfficial
- ChatGPT Images 2.5 launched; more than 3 billion a week reportedReport
- Big Technology cites 1.5 billion a weekReport
- ChatGPT Images 2.6 launched; more than 4 billion a week and more than 500 million a day reportedReport
- Sottiaux posts more than 4 billion a week, up from more than 3 billion 8 weeks earlierOfficial
The 700 million in 2025 was a figure for the first week right after launch, so its circumstances differ from those of later weekly figures.
Users grew from 1 billion to 1.2 billion in 2 months
Adam Fry, head of ChatGPT's consumer products, said on Nov. 2 (Nov. 3 Japan time) that weekly active users had grown from 1 billion to 1.2 billion over the past 2 months. The number of images grew from more than 3 billion to more than 4 billion over 8 weeks, and although the periods and scopes differ, users also increased over the same time. Whether the number of images per user rose cannot be confirmed because the scopes of the figures do not match.
The relationship to revenue has not been disclosed either. OpenAI announced on Oct. 5 that it would show ads next to image generation results and said it would begin testing in the U.S. in the latter half of October. The ads target users on the Free and Go plans, so more images could mean more opportunities to show ads. No figures showing the relationship between images and ad revenue have been found.
API revenue also cannot be estimated without a breakdown. API pricing for Images 2.6 per 1 million tokens is $5 for text input, $8 for image input and $30 for output. These prices apply only to images made through the API, and the API's share of the 4 billion has not been given. The cost of generation has not been confirmed either.
Google at more than 50 billion cumulative
The figures cannot simply be compared with those of competitors. Google said at its developer conference in May that more than 50 billion images had been created with its image generation model "Nano Banana." That is a cumulative figure, and a weekly count has not been confirmed. No material comparing the 2 companies' usage under the same conditions has been found.
New models continue to emerge in image generation. At OpenAI, the unannounced "aurora" appeared in the API list in October, and after the name "GPT Image 2.6" appeared in documentation, the company released 2.6 in November. On the evaluation site Arena, anonymous image models believed to be Microsoft's and "pumpkin," speculated to be the next model from China's HiDream, have also appeared.
