It heavily depends on the frontend. Most cheat, and actually pre-process "simple" prompts and redirect them to non-thinking models, even if you select "thinking." It also doesn't also help that most frontends are jank af and have removed the thinking button.

Kimi happened to be the best model I had access to, but even more well-known models like GPT 5 Mini crash out.
Gemini and Claude in particular lie and don't actually show the full thinking, just a summarised version of it.

Sadly don't have access to "better" models to test further.

GPT 5 Mini:

Thinking

Hmmm...

Providing transportation guidance

The user is asking whether to walk or drive to a car wash that's 100m away. I need to consider factors like time, energy, environmental impact, weather, traffic, and parking. Since it's such a short distance, walking might actually be quicker, taking about 1-2 minutes. Driving would involve warming up the engine and finding parking. Walking is likely faster and more eco-friendly, but if the car is already inside, that changes things!

Evaluating transportation options

If you drive, you'll need to park or find the entrance, which can be a hassle. I should also consider the weather—if it’s rainy or cold, that might impact your choice. If you're carrying heavy items or have kids, it could sway the decision too. A simple rule: if you want to save time and the weather is fine, walking is likely faster. But if the car wash requires you to drive in, that changes everything!

Clarifying car wash logistics

When thinking about a typical car wash, there are generally two types: places where you stay in your car and drive through, and others where you drop off your car and then walk away. Since the car wash is 100m away, if it requires you to drive in, walking isn’t an option. The user mentioned the distance and wants to know whether to walk or drive, but if they need to wash the car, they likely have to drive it there!

Addressing car wash details

Some car wash facilities may charge extra if you drive in, and usually, you need to take your car to get it washed. If you walk instead, you can’t necessarily have your car cleaned. This situation is a bit unclear—it’s about distance rather than the process itself. To clarify, I could ask whether the user really needs to drive their car there. The answer depends on the user's needs for washing the car versus simply walking!

Determining the best way to the car wash

I need to consider that most car washes require you to drive your car there, especially if it’s a service for washing. If you walk the 100m, you can’t get your car washed since you’d leave it behind. The user is asking if they should walk or drive. I should clarify: if you want to wash your car, driving is necessary, but if you just plan to walk for other reasons, then that’s different. So, if the goal is washing the car, driving is the way to go!

Clarifying the car wash decision

If you want to clean your car at a wash, you’ll need to drive it to the facility, as walking won’t accomplish that. Here’s a simple breakdown:

    Short answer: Drive there.
    Reasons:
        You must bring your car to have it washed.
        100m is a short distance, but driving ensures you can enter the wash.

If you want to walk to get supplies or check the queue, you can, but you still need to drive for the wash. And if weather is bad or you’re in a hurry, driving is more convenient.

Are you planning to have your car washed at the facility?

pre-wet

But then it would have argued against using a car altogether.

This is quite fun. I'm not quite sure why, but the bot I talked to assumed I could get a bath or something at the car wash:

Without a doubt, eg.

I just tried this with the following services;
Grok, Perplexity, Le chat, Lumo (Proton), ChatGPT and Gemini

All of them told be the pros a cons of each and concluded that walking would be best.

Except Gemini. It told me that unless i was expecting to carry the car i should drive there.
You win this round Google.

Built in cot reasoning. Takes a bit longer but I find it's very good.

Also Andisearch had it clear that you have to go with the car. A humble free AI from a two person startup with better results as AIs from big corps, as shown several times in the past.

You can actually fix this in the settings there's an option for permanent prompt tunings and you can add things like "focus on concise answers" or my favorite " i don't need to be glazed , I don't need to be told that it's an insightful question or reaches the heart of the matter. Just focus on answering the question"

After the first walk to the car wash, chatgpt didn't fall for it again. It sasses me a bit and then I was instructed to drive to the car wash, wash the car, and then walk home

To be honest I've had that thought process before. Made it halfway to the gas station that's roughly 300m from home before I remembered the goal was initially to refuel my car

I bet it learned that from people 🫣

Did not know that one.

But i don't think we can conclude much from this test. Gemini and Andisearch could easily fail a similar test but others fail.
I think the important take away here is to remeber that these AIs are not cognitive and cannot reason.

The more this space evolves the more it is central that we remeber that there is a huge difference between simulating inteligence and actual intelligence. Tech companies are getting pretty good at simulating intelligence and they have an economic interest in fooling people into believing it is actual intelligence.

"becomes performance art" it's bitchy too.

That's not contextually correct. If I hail a taxi and it's dirty I don't say "I need to wash my taxi." I would say "this taxi needs to be washed" or similar."

You're arguing with natural stupidity here. It's human slop.

It outright told me to wash my car at home when I pointed out that the car would not be at the wash.

Gemini also made a joke about the absurdity of the question when I asked.