← Back to Library

Gemini Exponential, Demis Hassabis' ‘Proto-AGI’ coming, but …

In the last 48 hours, we have had two major model releases and about 10 hours worth of interviews from top leaders about them. The insights of which I will try to condense into just 15 minutes or so. Because Gemini 3 flash is Google's attempt at finally convincing you to switch from chatbt or claude and the results look incredible. I'll go through them in a moment, but we have two co-founders of Google DeepMind.

Both seeing the LLM paradigm continuing on this exponential until a sketched out protoi model arrives in not too long. However, there are some problems with that vision and one result in particular I don't want you to miss. So, let's get started. Here are some of the raw numbers and bear in mind that the flash version of Gemini is the quick version, the one that can answer almost instantly.

You guys will know that all companies have a pro version of their models that typically take much much longer, minutes often to answer a question. I want you to notice the comparison with the model released 2 days ago, Gemini 3 Flash, with the state-of-the-art model as of June of this year, Gemini 2.5 Pro. Whether we're talking about academic reasoning, visual reasoning, scientific knowledge, coding, mathematics, the results aren't even that close. And this is for the dramatically quicker model.

For example, even without access to tools, the new Gemini 3 Flash roughly halves the error rate in one very difficult mathematics benchmark, AIM. Again, this is comparing Summer's Gemini 2.5 Pro at 88% to two days ago's Gemini 3 Flash at 95.2%. In fact, in almost any domain you can point to, from table and chart analysis, video analysis, or going off and being an agent, Gemini 3 Flash exceeds the previous huge model performance from the summer. You can, of course, optimize models for one particular set of benchmarks.

And we learned just this morning that Google did indeed apply a special type of post-training to optimize performance for software engineering. For those who code, you may be somewhat incredulous to see Gemini 3 Flash outperforming Gemini 3 Pro, the heavier model released just a few weeks ago. It would be very easy to get carried away with those results and say Chat GBT is doomed for consumers and Gemini, as Jim Kramer points out, is growing much faster. ...

Watch on YouTube →

Watch the full video by AI Explained on YouTube.