You’ve probably read about Sam Altman extolling the virtues of ChatGPT-5.4 Thinking as his “favorite model to talk to.” That’s great if you’re a Plus or Pro user, because you get access to that model. The rest of us on Free or Go accounts only have access to a version of ChatGPT based on GPT-5.3.
Until today, that is.
Today, OpenAI introduced GPT-5.4 mini and nano, describing them as “our most capable small models yet,” and the good news is that they’re available to Free and Go users in ChatGPT via a new “Thinking” option in the Plus (+) menu.
Let me think about that
You can’t choose GPT-5.4 mini or nano directly in the model selector in ChatGPT. Instead, you select the “Thinking” tool, and ChatGPT starts using them behind the scenes.
Even so, mini and nano bring a bit of GPT-5.4’s thinking magic to free users, and I’ve got to say, I’m impressed with what I’ve seen so far.
“Thinking” models work best when the answer isn’t obvious, requires multiple steps, or benefits from careful reasoning rather than quick recall. To test the new feature, I asked my free ChatGPT account to: “Plan a 5-day trip across Europe with a £800 budget, minimizing travel time and including 3 specific cities.”
I won’t paste the full travel plan here because it was quite long, but while the standard ChatGPT response was perfectly good, it was more of a bullet-point list.
By contrast, when I selected the Thinking tool, ChatGPT’s answer contained far more detail and gave me more insight into why it had chosen certain routes and places to visit. It left me feeling more confident that this was a genuinely well-thought-out plan.
In short, the answer felt more human.
Testing the new models
To test the models with a prompt involving multiple dimensions, I used: “I have £10,000, no coding experience, and 6 months. Design a realistic plan to generate income online, including risks, timelines, and fallback strategies.”
Again, the two plans it produced weren’t wildly different, but the execution was miles apart. Standard ChatGPT gave me a brief outline broken down into bullet points, while the Thinking version fleshed everything out and explained the reasoning behind its suggestions.
“I’ll give you a plan that balances speed, realism, and risk control, not hype,” it assured me. It also gave me two tracks to run in parallel — Track A for fastest income and Track B for building a scalable asset.
Both versions of ChatGPT included fallback strategies and risk analysis, but the Thinking version felt more solid and reliable.
Open and reliable
When OpenAI announced GPT-5.4 mini and nano, it was easy to assume they’d be compromises — lighter, cheaper versions of the “real” model designed mainly for background tasks and developers.
After actually using them, I’m not so sure.
After testing the new models across everyday prompts — writing, reasoning, and quick problem-solving — what stood out wasn’t what they couldn’t do. It was how often they felt almost indistinguishable from the full ChatGPT-5.4 Thinking model, just noticeably faster.
When I asked the full-fat ChatGPT-5.4 Thinking model on my Plus plan the same questions, it did a lot more of exactly that — thinking. Very slowly. Yes, it gave me more feedback on the decisions it was making, but it was essentially the same plan that GPT-5.4 mini or nano came up with. It just took longer to get there.
To be fair, the result was more personal and more detailed when I used a ChatGPT Plus account. But the fact that some measure of that thinking power is now available on a free ChatGPT account is seriously impressive.
I hope Plus users get access to them soon, too, because the speed boost feels well worth the slight trade-off in depth.
