ChatGPT's new Images 2.0 model is surprisingly good at generating text | TechCrunch

Open original

ChatGPT Images 2.0 uses a new model that overcomes diffusion-driven spelling errors and renders small text, icons, and UI elements cleanly at up to 2K resolution. It adds web search, multi-image prompts, and support for non-Latin scripts, with knowledge capped at December 2025. All ChatGPT and Codex users get access Tuesday, with paid tiers unlocking higher-quality outputs and an API.

  • ChatGPT's new Images 2.0 model renders text, icons, and UI elements accurately at up to 2K resolution, fixing the garbled-text problem that plagued earlier diffusion-based image generators
  • It adds "thinking" features like web search, multi-image generation per prompt, and self-verification of outputs, plus cleaner rendering of non-Latin scripts (Japanese, Korean, Hindi, Bengali)
  • Training data is capped at December 2025, so it may fumble very recent events
  • Rolling out Tuesday to all ChatGPT and Codex users, with paid tiers getting higher-quality output and a new gpt-image-2 API priced by resolution and fidelity

ChatGPT Images 2.0 now turns text prompts into restaurant-ready art. Two years ago, AI menus spat out “enchuita” and “burrto.” Today, asking for a Mexican menu yields plausible dishes—ceviche at $13.50 might raise an eyebrow, but no one’s spotting fake churros. Early diffusion models stumbled over spelling and small details because they treated text pixels as noise. Researchers have since tested autoregressive methods that predict images more like language models do. OpenAI won’t say which approach runs Images 2.0. The new model packs “thinking capabilities” that let it search the web, generate multiple images per prompt, and verify its own outputs. It handles marketing assets in assorted sizes and can stitch together multi-panel comics. Non-Latin scripts—Japanese, Korean, Hindi, Bengali—look cleaner, too. Its training data stops in December 2025, so very recent events may trip it up. Outputs can hit 2K resolution. According to OpenAI, Images 2.0 follows detailed instructions on text, icons, UI elements and complex layouts better than before. Generation takes a few minutes for complex scenes. Starting Tuesday, every ChatGPT and Codex user gets access; paid tiers unlock higher-quality results. OpenAI will launch a gpt-image-2 API with usage-based pricing based on resolution and output fidelity.

Ask about this article