OpenAI shipped ChatGPT Images 2.0 and the underlying gpt-image-2 API on April 21, 2026. It is a step-function image-generation upgrade and the biggest consumer image-gen release since GPT Image 1.5 in December 2025.
What’s new
- Native reasoning in the image model. The model thinks before it draws, handling multi-step compositional prompts (infographics, slides, annotated diagrams) that ripped GPT Image 1.5 and DALL-E 3.
- 4K resolution output at custom aspect ratios. Previous ceiling was 2K.
- Multi-image consistency across generations (character, product, brand).
- Multilingual text rendering across 12+ languages: English, Japanese, Korean, Chinese (Simplified and Traditional), Hindi, Bengali, and more. Previously the weak point for every major image model.
- Up to 8 distinct variants from a single prompt.
- Web search inside generation for Plus, Pro, and Business users (reference real-world branding, logos, product photography).
Availability
- ChatGPT Free: standard gpt-image-2 model, standard resolution, limited daily runs.
- ChatGPT Plus ($20/mo): thinking mode, longer reasoning runs, web search in generation.
- ChatGPT Pro ($200/mo): everything on Plus plus higher daily caps, priority generation.
- ChatGPT Business: thinking mode, commercial-use terms, team admin.
- gpt-image-2 API: available via OpenAI API and in Microsoft Foundry.
API pricing
| Quality tier | Resolution | Price per image |
|---|---|---|
| Low | 1024x768 | $0.01 |
| Medium | 2K | $0.08 |
| High | 4K | $0.41 |
Token-based pricing for long prompts: $8 per MTok input, $2 cached, $30 per MTok output (vs $32 output on gpt-image-1.5).
Competitive read
- Midjourney V7 remains the quality leader for artistic rendering. gpt-image-2 is closest yet, and its multilingual text rendering is genuinely better.
- Adobe Firefly loses its main commercial differentiator (brand-safe training data) as a pure feature play. Firefly’s IP indemnification and Adobe-stack integration still matter.
- Ideogram had owned text-in-image for a year. That moat is gone.
- Google Imagen 4 (built into Gemini) still leads on photoreal but trails on infographic and text-heavy generation.
The direct affordance shift: OpenAI bundles image gen into a $20/mo ChatGPT Plus subscription that already covers text, code, and voice. Midjourney’s $10/mo standalone plan is still cheaper, but for any user running both a chatbot and an image generator, Plus is now the default consolidation.
Open questions
- Video model timing. OpenAI’s Sora line remains Sora 2 (retired April 1) without public replacement. A Sora 3 release would parallel this image launch.
- Enterprise rollout. Business-tier indemnification terms on gpt-image-2 outputs have not been fully disclosed.
- Rate limits. OpenAI has not published hard caps for Plus/Pro tiers.
Related
Sources
Primary and corroborating references used for this news item.
- Introducing ChatGPT Images 2.0 - OpenAI
- ChatGPT's new Images 2.0 model is surprisingly good at generating text - TechCrunch (April 21, 2026)
- OpenAI's ChatGPT Images 2.0 is here - VentureBeat
- OpenAI unveils ChatGPT Images 2 - 9to5Mac
- Introducing OpenAI's gpt-image-2 in Microsoft Foundry
- GPT Image 2 Model - OpenAI API docs
Spotted an error or want to share your experience with OpenAI launches ChatGPT Images 2.0 and gpt-image-2 API: native reasoning, 4K output, multilingual text rendering?
Every tool page is re-verified on a recurring cycle, and corrections land faster when readers flag them directly. If you spot a stale fact, a missing capability, or have used OpenAI launches ChatGPT Images 2.0 and gpt-image-2 API: native reasoning, 4K output, multilingual text rendering and want to share what worked or didn't, the editorial desk reviews every message sent through this form.
Email editorial@aipedia.wiki