GPT-6 Launch Event Hosted by Tibo, the Cyber Godfather?
Editorial Desk · from Aofeisi Temple
2026-09-04 05:57:08
Source: QbitAI
As it turns out, this was fabricated by the new image-generation model GPT Image 2.5.
Although the model has not yet been officially released, some users have already received a new pop-up notification in ChatGPT.
Contrary to previous speculation about GPT Image 2.1, the version number has jumped straight to 2.5, suggesting a substantial upgrade.
So, just how powerful is GPT Image 2.5?
Based on the test results currently circulating online, it can already fabricate an official GPT-6 launch event and a Sam Altman all-hands letter, to the point where they are nearly indistinguishable from the real thing.
Here is a screenshot of a post allegedly published by Altman. Aside from the profile picture being incorrect, the text is virtually flawless.
Prompt: Take a realistic, phone-camera-style photo of a handwritten article. Write the article in pencil on 8.5×11-inch lined paper, with handwriting resembling Einstein’s. The content should be about the history of world currencies. Make sure the handwriting features clearly recognizable Einstein-like variations, and add a small coffee stain in the upper-right corner.
But what did Einstein’s handwriting actually look like? Here is a photo for comparison.
On the LMArena leaderboard, an anonymous model codenamed luna-lisa-alpha is suspected to be the new version of GPT Image. After some time spent testing it, users have identified the following characteristics:
- Extremely fast generation
- More photorealistic images
- Faces are less prone to distortion, even when generating multiple images of the same character
- Text embedded in images is less prone to distortion
- Improved on the noise issues found in GPT Image 2
Look at this tree—it’s really a tree. This house is really a house. This sheep is really a sheep. And this depth of field is really deep.
A direct comparison with GPT Image 2 makes the difference even more obvious.
For a while, the ChatGPT web version moved the image-generation feature up to the second spot in the main menu, just below “New chat,” but it has since been hidden again.
It looks like image generation will also become a flagship feature in the GPT-6 era.
References:
[1] https://x.com/chetaslua/status/2095556181012128005?s=20