Sponsored by
Hey everyone,
For three years, AI-generated images had one tell that gave them away instantly.
Look at any sign, poster, or label in the background and you would find gibberish. Letters that looked like letters but meant nothing. Words that almost existed. Typography that was purely decorative noise. Every AI image generator treated text as texture rather than language, and you could spot the fake from across the room.
On April 21, 2026, that era ended.
OpenAI released ChatGPT Images 2.0, built on a new model called GPT Image 2, with almost no announcement. No keynote. No hype cycle. Just a model page and a gallery. Within 12 hours it claimed the number one spot on every category of the Image Arena leaderboard by a margin of plus 242 points. The largest single lead ever recorded. By any model. Ever.
Today's primary sponsor provides financial support, contributing to my daily coffee expenses. I encourage you to explore their advertisement for further information.
What 2,000 SaaS Companies Reveal About Growth in 2026
Is your growth in-line with your peers in B2B SaaS & AI?
Benchmark yourself against actual billings data for Maxio’s 2000+ global customers, alongside firsthand company perspectives to understand how growth varied by company size, business model, and strategic focus.
Key takeaways from the report:
Average growth across 2,000 companies
Growth by revenue band
AI-led vs AI-enhanced. Who performed better?
Continue to Post
What Makes This Model Different
It thinks before it draws
Every image model before this one started generating pixels immediately. Images 2.0 integrates OpenAI's reasoning architecture, so the model plans layout, searches the web for reference, checks its own output, and refines before delivering. It can generate up to eight images from a single prompt simultaneously and present you with the best versions based on its own quality assessment.
This is not a better generator. It is the first image model with a brain.
The Text Problem Is Solved
Why every designer and marketer needs to pay attention
GPT Image 2 delivers approximately 99% text accuracy in any language or script. Posters, menus, infographics, marketing materials, conference banners, social graphics. Every word correct. Every date right. Every name spelled properly.
When TechCrunch tested the model with a Mexican restaurant menu prompt, the result was something that could immediately be used in a restaurant without customers noticing anything was off.
That workflow used to take two to three hours in Canva or Photoshop. It now takes 30 seconds.
Eight Things Worth Trying This Week
YouTube thumbnails with bold, correctly-spelled text overlays
Social media graphics with brand colours and call to action text
Product mockups in lifestyle settings without a photo shoot
Full infographics from a single detailed prompt
Email headers and marketing banners ready to use directly
Presentation slide visuals with eight variations generated at once
Educational diagrams with correct labels and captions
All of the above in Japanese, Korean, Chinese, Hindi, or Bengali
That last point matters enormously. For the first time, AI image generation is genuinely useful for billions of people whose languages were previously mangled by every model.
Today's Secondary sponsor provides financial support, contributing to my daily coffee expenses. I encourage you to explore their advertisement for further information.
[Webinar] Stop babysitting your coding agents
MCPs give your agents access to information, not understanding. The teams pulling ahead are using a context engine to give agents the right context for every task, so they stay on track without the set up tax or the correction loops. Join live on May 6 (FREE) to see how.
Let's finish the post
Who Gets What
Access Level | What You Get |
|---|---|
Free | Core quality upgrade, text accuracy, better instruction following |
Plus and Pro | Web search before generating, 8-image batching, self-verification, 2K resolution |
API | Up to 4K resolution in beta, cheaper per image than GPT Image 1.5 |
One More Thing
DALL-E 2 and DALL-E 3 are both being retired on May 12, 2026. The model that launched the AI image era is being switched off. GPT Image 2 is the replacement across ChatGPT and the entire API.
DALL-E was a generator. GPT Image 2 is a designer. That distinction is the whole story.
Open ChatGPT today and ask for an image. You are already using it.
Catch you next time,
— Raja Tahoor Ahmad
P.S. Share this with a designer or marketer who still thinks AI image generation is too unreliable for real work. They should try it this week.