Sam Page
← All thinking

Testing the BRAND NEW Gemini 3 against Chat GPT 5.1 on the most CRITICAL task we all…

Emerging tech and AI3 min readSam Page

Testing the BRAND NEW Gemini 3 against Chat GPT 5.1 on the most CRITICAL task we all need support with... buying a Christmas gift for our loved ones.

Google has just released Gemini 3 and in general intelligence tests, it out performs all other models... JUST. Scoring 73%, versus GPT-5.1 at 70%.

On "Humanity's Last Test" it does much better, scoring 37.2% versus GPT-5.1 26.5%.

Perhaps it's most impressive feature is a 1m token context window. In essence it's short term memory. GPT-5.1 has a much smaller 128k window in comparison. This means Gemini should be much better at processing vast amounts of information, and remembering context in a conversation for far longer. And it's multi-modal capabilities are supposed to be impressive (text, image and video etc).

But the real test... which model can best help me find a gift for my wife, Gemma Page CPCC, this Christmas.

My prompt (deliberately basic at this stage, doesn't make use of the huge context window):

I'm looking for a Christmas Gift for my wife, Gemma. We have two children and a dog. She's a life coach for working mums. We live in Kingston upon Thames. She loves candles, but lots of people buy them for her. I'm looking to spend up to £100.

Gemini 3 took more time to respond, but provided back 3 clear options, each which had 3 components to them, well through through and explained. It had considered the context and specifically researched local options, that Gemma could go to, along with thoughtful products to sit alongside them. It named the local businesses, and provided actual products, and split the budget down really well. It also gave a really helpful summary and table at the end.

GPT-5.1 gave a fast response, and provided back 6 separate options. Each option is well considered, with the reasoning for the suggestion. In places it has recommended a company or product, although it fell short on actually suggesting local places like Gemini did. At the end it recommended how three could be combined to use the total budget. It didn't theme the bundle, but rather created a mix of items.

It's a simple test, and of course I could have prompted both to get a more structured output, but my net conclusion is that Gemini 3 did more research, provided better and more accurate recommendations, and packaged the gifts up together in an intelligent way that built a theme.

I then asked each model to create an image for me to put on the front of a Christmas Card to reflect the first option it recommended.

Gemini 3 was incredibly quick, and I think has done an extremely good job of creating a realistic and tailored image.

GPT-5.1 has sligtly mis-interpreted the request and got the context of the wording wrong. Which is pretty typical of it's image generation capablity.

Gemini's the clear winner on this one for me.

Merry Christmas Gemma Page CPCC, courtesy of Google's Gemini 3. 😉

More tests to follow...

First published on LinkedIn.

More on emerging tech and AI

One email, every Friday.

What happened this week, what I'm thinking about, and one takeaway for yours.

GEThttps://sam-page.com/thinking/testing-the-brand-new-gemini-3-against-chat-gpt-5-1-on-the-most-critical-task-we.mdtext/markdown; charset=utf-8

---
title: "Testing the BRAND NEW Gemini 3 against Chat GPT 5.1 on the most CRITICAL task we all…"
url: https://sam-page.com/thinking/testing-the-brand-new-gemini-3-against-chat-gpt-5-1-on-the-most-critical-task-we
date: 2025-11-20
author: Sam Page
type: Post or article
categories: [Emerging tech and AI]
reading_time: 3 min
summary: "Testing the BRAND NEW Gemini 3 against Chat GPT 5.1 on the most CRITICAL task we all need support with... buying a Christmas gift for our loved ones."
first_published: https://www.linkedin.com/feed/update/urn%3Ali%3AugcPost%3A7397173559200174080
---

# Testing the BRAND NEW Gemini 3 against Chat GPT 5.1 on the most CRITICAL task we all…

Testing the BRAND NEW Gemini 3 against Chat GPT 5.1 on the most CRITICAL task we all need support with... buying a Christmas gift for our loved ones.

![](https://sam-page.com/img/thinking/testing-the-brand-new-gemini-3-against-chat-gpt-5-1-on-the-most-critic-li1.webp)

![](https://sam-page.com/img/thinking/testing-the-brand-new-gemini-3-against-chat-gpt-5-1-on-the-most-critic-li2.webp)

Google has just released Gemini 3 and in general intelligence tests, it out performs all other models... JUST.  Scoring 73%, versus GPT-5.1 at 70%.

On "Humanity's Last Test" it does much better, scoring 37.2% versus GPT-5.1 26.5%.

Perhaps it's most impressive feature is a 1m token context window. In essence it's short term memory. GPT-5.1 has a much smaller 128k window in comparison. This means Gemini should be much better at processing vast amounts of information, and remembering context in a conversation for far longer. And it's multi-modal capabilities are supposed to be impressive (text, image and video etc).

But the real test... which model can best help me find a gift for my wife, Gemma Page CPCC, this Christmas.

My prompt (deliberately basic at this stage, doesn't make use of the huge context window):

---  
I'm looking for a Christmas Gift for my wife, Gemma. We have two children and a dog. She's a life coach for working mums. We live in Kingston upon Thames. She loves candles, but lots of people buy them for her. I'm looking to spend up to £100.  
---

Gemini 3 took more time to respond, but provided back 3 clear options, each which had 3 components to them, well through through and explained. It had considered the context and specifically researched local options, that Gemma could go to, along with thoughtful products to sit alongside them. It named the local businesses, and provided actual products, and split the budget down really well. It also gave a really helpful summary and table at the end.

GPT-5.1 gave a fast response, and provided back 6 separate options. Each option is well considered, with the reasoning for the suggestion. In places it has recommended a company or product, although it fell short on actually suggesting local places like Gemini did. At the end it recommended how three could be combined to use the total budget. It didn't theme the bundle, but rather created a mix of items.

It's a simple test, and of course I could have prompted both to get a more structured output, but my net conclusion is that Gemini 3 did more research, provided better and more accurate recommendations, and packaged the gifts up together in an intelligent way that built a theme.

I then asked each model to create an image for me to put on the front of a Christmas Card to reflect the first option it recommended.

Gemini 3 was incredibly quick, and I think has done an extremely good job of creating a realistic and tailored image.

GPT-5.1 has sligtly mis-interpreted the request and got the context of the wording wrong. Which is pretty typical of it's image generation capablity.

Gemini's the clear winner on this one for me.

Merry Christmas Gemma Page CPCC, courtesy of Google's Gemini 3. 😉

More tests to follow...