Testing Google’s New Gemini 2.5 Pro Model

+

Google recently dropped its most advanced model yet, 2.5 Pro. This morning, I put it through 3 real world tests to see if Google can beat the best model I’ve used so far: Grok 3.

I asked it 3 questions I actually need the answer to. These are the same questions I posed to Grok 3 recently.

Let’s get started!

1) Optimize My Stock Portfolio. I recently had Grok help me fine tune my index fund portfolio. But maybe Gemini can do better.

Here’s the prompt:

“Assume I have a portfolio of index funds. My goal with this portfolio is long term growth. How should I allocate this portfolio across different types of index funds? Consider options like US index funds, foreign stock index funds, etc. I am 39 years old, so my time horizon is long. My risk tolerance is high. What would be the best allocation, given all this information? Use the best research you can find to support your answer.”

Let’s see what Gemini says:

Gemini’s response was smart and well-reasoned. Like Grok, it recommends a portfolio of 90-100% stocks with a split between domestic and foreign.

But there was one big problem with Gemini’s answer: it had no sources.

Gemini mentioned that Modern Portfolio Theory supports its answer. But there’s no link to any research and no way to verify its claim.

Grok 3, by contrast, cited over 40 high quality sources. I’m giving this round to Grok.

2) Find Me Some Startups. I’m really interested in startups that could help us with our baby bust in America. So, I set Gemini to work on it.

Here’s the prompt I used:

“What are the most interesting startups at pre-seed stage working on solutions to increase fertility? Consider startups to lower costs and improve effectiveness of IVF, and also startups to improve egg health for older women, among other possible ways to boost fertility. Please only show me startups that have raised $750,000 or less in funding.”

Here’s what Gemini came up with:

Gemini’s answer was useless. It gave me a startup, Gameto, that has raised $73 million in funding. This is a clear violation of the instructions I gave. Similarly, Alife has raised over $30 million.

Then, it gave me some general ideas of types of startups to look at. This isn’t useful either — I asked for specific companies.

Grok, on the other hand, gave me some actual early stage startups to research. Once again, the round goes to Grok.

3) Making My Meetings Better. Finally, I asked Gemini to help me with a problem I struggle with daily: how to do my best when I meet with founders.

I want to be present and helpful. I want to ask good questions and learn a lot about the founder’s vision.

Here’s the prompt I used:

“As an angel investor, I meet with a lot of startup founders. I want to do the best job I can in those meetings. What are some tips to perform better, be more helpful, and learn more about the startups I meet with?”

Let’s see what Gemini advises…

Gemini’s advice is spot on. It tells me to be sure I’m on time and to offer constructive feedback when I pass on a company. Those are things I already try to do, but it’s good to have that emphasized.

Gemini’s answer was just as good as Grok’s here. So, I’ll call this round a tie.

Wrap-Up

Grok 3 beat Gemini 2.5 handily, winning 2 out of 3 rounds.

If Google wants to dethrone the leaders like Grok and ChatGPT, they’re going to have to do better than this. I expect more out of a $2 trillion company.

This model is actually worse than their existing Deep Research model, which gives precise and well sourced responses. With Gemini 2.5, Google is actually going backwards.

If you need answers to your questions or want to build AI into your product, Gemini isn’t it. Grok 3 is a better choice.

Have a great weekend everybody!

Testing Google’s New Gemini 2.5 Pro Model

Share this:

One response to “Testing Google’s New Gemini 2.5 Pro Model”

Leave a comment Cancel reply