Learning Timeline
Key Insights
Advantages of 'Heavy Thinking Effort'
'Heavy' mode takes longer to process than 'Standard', but it delivers significantly higher quality analysis and is ideal for tasks requiring high data precision.
Real-Time Context Update Feature
You don't need to wait for the AI to finish its response before adding more info. GPT 5.4 lets you 'interfere' or add context mid-process, and the AI will adapt its answer based on the new input immediately.
Reduced Hallucination
Model 5.4 has a 33% reduction in 'hallucination' compared to version 5.2, making it more reliable for tasks that require authentic citations.
Prompts
AI Hallucination Analysis Research
Target:
ChatGPT (GPT 5.4)
Perform a research analysis on whether consumer AI products have reduced hallucination over time. I need the output in three specific sections: 1. An upfront research plan, 2. The findings with citations, and 3. A final checklist. Start by looking for 10 sources.
Follow-Up Prompt During 'Thinking'
Target:
ChatGPT (GPT 5.4)
Actually, give me 15 sources instead of 10 and ensure they are all from the last 12 months.
Step by Step
How to Perform Deep Research with GPT 5.4 Thinking Mode
- Open the ChatGPT interface and ensure you are using the latest model (GPT 5.4).
- Locate the 'Thinking effort' menu situated near the message input bar.
- Click the menu and switch the setting from 'Standard' (default) to 'Heavy' for more in-depth and complex analysis.
- Enter your research prompt and specify your desired output format (e.g., Research Plan, Findings with Citations, and Checklist).
- Click the 'Send' button to initiate the 'Thinking' and 'Research' process.
- While the AI is 'Thinking', you can type additional instructions or context (follow-up) in the same chat bar.
- Click 'Send' on the follow-up prompt to redirect the research in real-time without needing to pause or start a new chat.
- Wait for the AI to finish generating a comprehensive output based on your latest instructions.
