GPT-Image-2: 10,217 AI-Generated Images from Twitter in First Week Post-Launch

10,217 images confirmed to be generated by the GPT-Image-2.0 model, sourced from public Twitter/X posts. The dataset was collected by Scam-AI during the first week after the model's release, from April 21 to April 28, 2026. It supports research into the real-world use and characteristics of a newly launched generative AI model.

Use Cases

• Analyze the thematic content and visual styles of early-adopter generated images based on the social media source.
• Study the real-world application patterns of a new AI model based on the first-week post-launch collection window.
• Benchmark detection algorithms for AI-generated imagery based on a confirmed set of model outputs.
• Investigate potential biases or failure modes in a generative model's early public use based on user-shared content.

Strengths

• Contains 10,217 confirmed GPT-Image-2.0 outputs, providing a substantial sample.
• Has a precisely defined collection window (April 21–28, 2026) capturing immediate post-launch activity.
• Images are sourced from a real-world platform (Twitter/X), reflecting organic user behavior.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset's scope is limited to a single platform and a one-week period, which may not be representative of longer-term use.
• Description metadata is limited; actual data quality and image attributes require manual inspection after download.

Provenance

Source

Twitter/X posts.

Collection Method

Collected from public social media posts.

Time Range

April 21, 2026 – April 28, 2026

Freshness

Last updated 2026-05-20 03:33:20; freshness should be verified.