Source record

@tjrobertson52 TikTok profile avatar

@tjrobertson52 source record

What Gemini 3 Flash Means for Search – Google's racing to replace its own search before ChatGPT does. The problem? When it doesn't know something, it gives a wrong answer 90% of the time instead of saying "I don't know." #GoogleAI #SEO #TechNews #AISearch

2025-12-21published
tiktokplatform
2public insights

Topics Mentioned

Public Evidence Excerpt

i

Gemini 3 Flash came out this week and I wanna talk about why this is actually a big deal and something worth paying attention to. So you might know Gemini 3 came out a while back. Gemini 3 Flash is the smaller, faster version and all the large language models have one of these nano or flash versions. It's the model you want to use if you need a quick answer and don't want to spend a lot of tokens. So why would we care about the faster, dumber version of G P t 3? Well, 1, because it turns out it's not nearly as dumb as we assumed it would be. And 2, because it tells us a lot about Google's plan for search. When you look at the most popular benchmarks, Gemini 3 Flash is actually very close to Gemini 3 Pro. It's even beating Gemini 3 Pro on a few benchmarks. Now there is one major flaw with these benchmarks, but I'll get back to that. I wanna talk about why this is really worth paying atten

Related Passages

i

These are public discovery snippets linked to the same source record. A snippet can end early when the public page keeps only short evidence context.

Gemini 3 Flash came out this week and I wanna talk about why this is actually a big deal and something worth paying attention to. So you might know Gemini 3 came out a while back. Gemini 3 Flash is the smaller, faster version and all the large language models have one of these nano or flash versions. It's the model you want to use if you need a quick answer and don't want to spend a lot of tokens. So why would we care about the faster, dumber version of G P t 3? Well, 1, because it turns out it's not nearly as d...
rtain question correct. However, there's no penalty for an incorrect answer. So while Gemini 3 Flash does get more correct answers than almost any other model, it's also one of the models least likely to say I don't know, which means it still has a high hallucination rate. In fact, over 90% of the time that it doesn't know the answer, it will still give a confident answer. Google is well aware of this issue, which is why they continue to experiment with hybrid solutions, experiments like Google's web view, where...

Public Insight Cards

i

Google may move toward hybrid search experiences mixing AI responses with traditional results to reduce hallucination risk.

AI search / hybrid results · asserts

rtain question correct. However, there's no penalty for an incorrect answer. So while Gemini 3 Flash does get more correct answers than almost any other model, it's also one of the models least likely to say I don't know, which means it still has a high hallucination rate. In fact, over 90% of the time that it doesn't know the answer, it will still give a...

Open

Fast smaller models like Gemini Flash may answer confidently even when wrong, making hallucinations a major constraint for AI search.

Risk/avoid / AI hallucinations · asserts

Gemini 3 Flash came out this week and I wanna talk about why this is actually a big deal and something worth paying attention to. So you might know Gemini 3 came out a while back. Gemini 3 Flash is the smaller, faster version and all the large language models have one of these nano or flash versions. It's the model you want to use if you need a quick answ...

Open