OpenAI rolls out GPT-5.4 Thinking in ChatGPT with new control over AI responses
OpenAI has begun rolling out GPT-5.4 in ChatGPT, introducing a new “GPT-5.4 Thinking” mode that allows users to interrupt the model while it is generating a response and steer the output before the final answer is produced.
Nick Turley, Head of ChatGPT at OpenAI, shared the update on LinkedIn, outlining how the new model expands reasoning capabilities and introduces new controls designed for professional work and complex tasks.
Turley wrote on LinkedIn: “GPT-5.4 Thinking is rolling out to ChatGPT. You can now interrupt it before it produces the final answer.”
He added that the feature allows users to guide the model’s reasoning while it is still generating a response: “That means you can steer the response while it’s still working instead of needing multiple back-and-forth turns.”
Interruptible reasoning changes how users interact with AI
The GPT-5.4 Thinking mode introduces a planning step where the system can outline how it intends to approach a task. Users can review that plan and adjust direction before the model produces a final output.
OpenAI said the change is intended to reduce the need for repeated prompt refinement, allowing users to guide responses earlier in the process rather than after an answer is completed.
Turley also highlighted improvements to research and long-context reasoning. He wrote: “We also improved deep web research and long-context reasoning, so answers to highly specific questions arrive faster and stay focused on what you’re trying to do.”
Model designed for professional and developer workflows
OpenAI said GPT-5.4 combines recent advances in reasoning, coding, and agent-based workflows into a single frontier model.
The company said the system incorporates coding capabilities developed through earlier Codex models while improving how the model interacts with documents, spreadsheets, presentations, and other professional tools.
GPT-5.4 is also being made available through the OpenAI API and Codex, where it introduces expanded computer-use capabilities designed for agent systems that can operate across software environments and multi-step workflows.
The model supports context windows of up to one million tokens, allowing it to process larger volumes of information across complex tasks.
OpenAI said GPT-5.4 also improves token efficiency compared with earlier versions, meaning the system can complete reasoning tasks using fewer tokens while maintaining performance.
The update reflects a broader shift in AI development, as models increasingly move beyond conversational use cases toward supporting structured knowledge work, software development, and complex professional tasks.
ETIH Innovation Awards 2026
The ETIH Innovation Awards 2026 are now open and recognize education technology organizations delivering measurable impact across K–12, higher education, and lifelong learning. The awards are open to entries from the UK, the Americas, and internationally, with submissions assessed on evidence of outcomes and real-world application.
