GPT-5.4 Thinking offers stronger reasoning but struggles with images

OpenAI's GPT-5.4 Thinking is presented as a more cognitively prepared model, and a reviewer tested it via the $20-per-month ChatGPT Plus plan to see how it handles deeper challenges. In text-based tasks the model delivers deeper analysis and generally strong reasoning; most responses were thoughtful and free of obvious hallucinations.

However, it sometimes answered questions different from those asked, and its formatting habits—long numbered lists and odd labels—proved awkward. Specific tests highlighted the split performance. An aircraft-carrier prompt produced a solid engineering analysis that explained why four downward-facing turbo‑propellers are a poor solution, yet the image outputs kept the same incorrect propeller orientation and returned gibberish labels.

A Boston tech-and-history itinerary was useful but initially presented as an unwieldy numbered list before being reformatted; budget and lodging suggestions were practical.

gpt-5.4, openai, chatgpt plus, multimodal, reasoning, hallucinations, image generation, propeller orientation, aircraft carrier, boston itinerary