Wild GPT-image-2 use cases

Nice article. From DALLE to the GPT Image series (1, 1.5, and now 2), it’s crazy how much things have improved on the image model for OpenAI. Gpt-Image in the editing benchmarks of my GenAI Showdown shot up from scoring 3/12 to 10/12.

The example on your site where it successfully creates/integrates valid QR codes and barcodes is pretty wild.

I’m curious how far this can go with extended thinking. From what I can tell - this automated web search to extends its knowledge beyond the training data cutoff point of Dec 2025 seems limited to the ChatGPT interface.

I used the OpenAI API to call GPT Image 2 directly and asked it to create an infographic based on the results of the April 2026 London Marathon. It presented the image as if it were using up-to-date data but it clearly pulled the records from 2025 missing the fact that runner Sawe broke the world record at 1:59:30.

Looks like you’re using the API directly, you basically have to handle the web search yourself, e.g. wrap the tool usage in your own retrieval or web search pipeline and then inject that information into the prompt. Either way, it’s still incredibly cool.

https://genai-showdown.specr.net/image-editing?models=o4,g15...

Just a heads up: this article is effectively paywalled, or at least gated behind having an account. I think you’re going to lose a lot of potential readers (particularly on HN) if you stick to Medium.