ChatGPT just got another brain boost with GPT-5.4 Thinking — and it’s built for bigger, more complex tasks
The new model explains its plan before answering, letting users correct its approach mid-process.
OpenAI has introduced a significant architectural shift with the launch of ChatGPT 5.4 Thinking, a specialized model powered by GPT-5.4 designed for complex reasoning. This new offering, available as an optional mode for paying subscribers (Plus, Team, and Pro), represents a move away from a one-size-fits-all AI toward a layered system. It joins the existing ChatGPT 5.3 Instant, which remains the default for fast, conversational interactions. The core innovation is the model's 'thinking out loud' capability, where it previews its step-by-step plan for tackling a user's prompt, allowing for real-time correction before the final answer is generated.
Technically, GPT-5.4 excels in professional domains, matching or exceeding human performance in tasks like financial spreadsheet modeling, document analysis, and presentation design. It features improved visual understanding for interpreting charts and complex documents, and maintains stronger conversational context for multi-step problems. This release signals OpenAI's focused push to make ChatGPT indispensable for professional workflows, moving beyond chat into a collaborative reasoning agent. The model's ability to be guided mid-process could fundamentally change how professionals use AI for planning, analysis, and complex project execution.
- GPT-5.4 Thinking explains its reasoning plan before answering, allowing user intervention and correction.
- The model matched or exceeded human professionals in tests for document creation, spreadsheet analysis, and presentation design.
- It's an optional mode for paying subscribers, while ChatGPT 5.3 Instant remains the default for everyday speed.
Why It Matters
Transforms AI from a black-box answer generator into a collaborative, steerable partner for complex professional work.