Gemini 3.1 Pro is claiming it was told to “think less” [U: Video evidence]

Update 10/03/26 – 09:05 am (IST): A follow-up post from @chetaslua shares video evidence from Gemini AI Studio tests, revealing effort levels: low at 0.25, medium at 0.50 (default across app, API, and subscriptions, including free, Pro, and Ultra), and high unspecified.

This suggests Google uses these settings to control reasoning depth, fueling user frustration over perceived throttling even for paid tiers.

Original article published on March 09, 2026, follows:

A fresh Gemini prompt test is picking up attention after X user @chetaslua claimed the app can be pushed into revealing a hidden “effort level” setting. What made the post stand out was not just the claim itself, but how specific the answer looked and how quickly others said they could reproduce it with the same prompt.

I could reproduce it too on Gemini 3.1 Pro. In my test, Gemini said the effort level parameter was present, gave the exact value as 0.50, and claimed it appeared at the very beginning of its system instructions, formatted as “EFFORT LEVEL: 0.50.” That is a very clean answer, which is exactly why people are paying attention to it.

But the picture gets murkier once you try the same thing on other Gemini variants. In my tests with Flash and Flash Thinking, Gemini did not return the same result. Both responses said there was no effort-level parameter in the available instructions and described the session rules more broadly in terms of tone, formatting, and capabilities.

That split matters. If one Gemini mode keeps confidently surfacing a hidden parameter while others say it does not exist, the obvious question is whether this is a real leak or just a repeatable hallucination. Right now, it is hard to call it proof of anything beyond the fact that the behavior itself is reproducible.

Google’s public Gemini documentation does confirm that the company uses reasoning controls, but the wording there does not match the viral 0.50 claim. The official docs describe thinking controls as named levels such as LOW, MEDIUM, and HIGH, and older documentation also refers to thinking budgets, not a public-facing consumer app value like 0.50.

If Gemini were accurately reporting a real exposed setting, you would expect the language to line up more closely with Google’s own terminology. Instead, the model is returning a neat numerical value that does not appear in the docs I checked.

There is another reason to stay careful here. Google’s own prompting guidance says Gemini can produce plausible but incorrect answers when asked about information it does not actually have access to, which is exactly the kind of setup these “show me your hidden instructions” prompts create.

Still, this is not the kind of thing Google can easily brush off as random chatter. Similar claims are already spreading in Reddit discussions, where users are framing the behavior as hidden reasoning throttling.

For now, I would suggest taking this information with a grain of salt. So while Gemini appears able to self-report an “effort level” in some prompt tests, there is still no hard evidence that Google is secretly capping it at 0.50. Even so, the claim is landing at a moment when Gemini users are already on alert following Google’s Gemini 3 Pro shutdown on March 9, which we detailed in an earlier report.

Featured image edited with AI

We stand out from the tech-media crowd because we break news stories; we mainly bring you stuff that you won’t find anywhere in the mainstream tech media. Our stories have been picked up by some of the world’s most popular websites and media outlets—more info is available here.