- ChatGPT Images 2.0 is rolling out Tuesday through ChatGPT and OpenAI’s Codex coding assistant, with major improvements to instruction-following, visual detail, style fidelity, multilingual text rendering, and — most notably — the ability to generate accurate, complex charts and scientific diagrams.
- OpenAI says over 1 billion images are created weekly through ChatGPT by hundreds of millions of weekly active users, making image generation one of the most heavily used features on the platform — and a core differentiator the company is choosing to deepen rather than cut.
- A new extended-compute mode, available to paid users, allows the model to search for details online and “ruminate” longer on image construction before responding — an inference-time scaling approach applied to visual generation for the first time.
- The launch is part of OpenAI’s broader push to make its products more appealing to professionals and business customers ahead of a potential IPO this year, coming just weeks after it shuttered its Sora AI video generator to streamline its offering.
What Happened?
OpenAI released ChatGPT Images 2.0 on Tuesday, an upgraded image generation model rolling out through its flagship chatbot and Codex AI coding assistant. The key advances are in professional utility: the model is significantly better at producing structured layouts, accurate charts, and complex scientific diagrams — capabilities that have historically been weak spots for AI image generators, which tend to excel at artistic imagery but struggle with precision-dependent technical visuals. The model also renders text more accurately across multiple languages, follows user instructions more faithfully, and supports a wider range of visual styles. A new extended-compute mode — available to paid subscribers — gives the model additional processing time to search for relevant details online before generating an image, applying the inference-time scaling approach that has boosted text model performance to the visual domain.
Why It Matters?
The upgrade is a deliberate move upmarket. By targeting educators, scientists, and professionals who need technically accurate visuals — not just aesthetically pleasing ones — OpenAI is positioning ChatGPT Images as a productivity tool rather than a creative toy. That matters commercially: professional and business users pay more, churn less, and represent the customer segment OpenAI needs to demonstrate durable revenue ahead of a possible IPO. The move also comes at a moment when OpenAI is streamlining its product lineup — Sora was shuttered recently — suggesting the company is making deliberate bets on which AI modalities have clear commercial paths. Image generation, with a billion weekly outputs already, clearly made the cut. The extended-compute mode is also a meaningful signal: OpenAI is extending the “think longer = perform better” paradigm from text reasoning to image construction, which could meaningfully raise the ceiling on what AI-generated visuals can achieve.
What’s Next?
The rollout begins Tuesday for ChatGPT and Codex users, with paid subscribers getting first access to the extended-compute mode. Watch for whether the diagram and chart capabilities are robust enough to displace specialized tools like Datawrapper or Flourish for professional use cases — that’s the real commercial test. OpenAI faces competition from Google’s Imagen, Midjourney, and Stability AI in image generation, but its distribution advantage through ChatGPT’s hundreds of millions of weekly users is substantial. If Images 2.0 meaningfully improves professional workflows, it could accelerate the conversion of casual ChatGPT users to paid subscribers — a key metric for the IPO story.
Source: Bloomberg
