Key Points
- OpenAI released a new ChatGPT image model (1.5) with clearer text rendering.
- The model still makes directional errors in detailed graphics like a water‑cycle infographic.
- Google’s Nano Banana consistently produces higher‑quality realistic images faster.
- Free usage limits on Google’s platform last longer than those on OpenAI’s service.
- ChatGPT excels at cartoon‑style and highly creative prompts.
- The update appears to be a response to Google’s recent advances in AI image generation.
OpenAI’s Updated Image Model
OpenAI introduced a new iteration of its ChatGPT image generator, often referred to as the 1.5 model. The upgrade brings noticeable improvements in text legibility and overall image quality compared with earlier versions. Testers observed that the model can now produce clearer textual elements in generated images, a step forward from the prior generation where text was often difficult to read.
Persistent Accuracy Issues
Despite these gains, the model still exhibits errors in precise visual details. In a practical test involving a water‑cycle infographic, the generated arrows pointed in the wrong direction, giving the impression that rain could rise from the ground. Although the latest model rendered the text more clearly, the directional mistakes persisted, indicating that the system’s grasp of specific scientific diagrams remains limited.
Performance Compared to Google’s Nano Banana
Google’s Nano Banana, part of its Gemini suite, consistently outperformed the new ChatGPT model in head‑to‑head comparisons. Nano Banana produced more realistic images with higher fidelity and did so more quickly. Users also reported that the free usage limits on Google’s platform lasted longer than those on OpenAI’s service, reducing interruptions during testing.
Strengths in Creative, Non‑Realistic Prompts
The ChatGPT image generator continues to excel at generating stylized, cartoon‑like visuals. Its ability to handle imaginative, less realistic prompts remains a distinctive advantage, even as the model’s realistic rendering capabilities lag behind Google’s offering.
Implications for the AI Image Race
The rollout of the updated ChatGPT image model appears to be a response to Google’s rapid advancements in generative AI. While OpenAI has closed some gaps—particularly in text clarity—the overall quality, speed, and realism of its images still fall short of Nano Banana’s performance. Observers suggest that OpenAI may continue to iterate quickly to narrow this divide, but the current comparison underscores a clear lead for Google in the competitive landscape of AI‑driven image creation.
Source: cnet.com