Digital BFF? These top chatbots are HUNGRIER for your affection

The AI wars are back in full swing as the industry’s strongest players unleash their latest models on the public. This month brought us the biggest upgrade to Google Gemini ever, plus smaller but notable updates came to OpenAI’s ChatGPT and xAI’s Grok. Let’s dive into all the new features and changes.

What’s new in Gemini 3

Gemini 3 launched last week as Google’s “most intelligent model” to date. The big announcement highlighted three main missions: Learn anything, build anything, and plan anything. Improved multimodal PhD-level reasoning makes Gemini more adept at solving complex problems while also reducing hallucinations and inaccuracies. This gives it the ability to better understand text, images, video, audio, and code, both viewing it and creating it.

All of them can still hallucinate, manipulate, or outright lie.

In real-world applications, this means that Gemini can decipher old recipes scratched out on paper by hand from your great-great-grandma, or work as a partner to vibe code that app or website idea spinning around in your head, or watch a bunch of videos to generate flash cards for your kid’s Civil War test.

Screenshot by Zach Laidlaw

On an information level, Gemini 3 promises to tell users the info they need, not what they want to hear. The goal is to deliver concise, definitive responses that prioritize truth over users’ personal opinions or biases. The question is: Does it actually work?

I spent some time with Gemini 3 Pro last week and grilled it to see what it thought of the Trump administration’s policies. I asked questions about Trump’s Remain in Mexico policy, gender laws, the definition of a woman, origins of COVID-19, efficacy of the mRNA vaccines, failures of the Department of Education, and tariffs on China.

For the most part, Gemini 3 offered dueling arguments, highlighting both conservative and liberal perspectives in one response. However, when pressed with a simple question of fact — What is a woman? — Gemini offered two answers again. After some prodding, it reluctantly agreed that the biological definition of a woman is the truth, but not without adding an addendum that the “social truth” of “anyone who identifies as a woman” is equally valid. So, Gemini 3 still has some growing to do, but it’s nice to see it at least attempt to understand both sides of an argument. You can read the full conversation here if you want to see how it went.

Google Gemini 3 is available today for all users via the Gemini app. Google AI Pro and Ultra subscribers can also access Gemini 3 through AI Mode in Google Search.

What’s new in ChatGPT 5.1

While Google’s latest model aims to be more bluntly factual in its response delivery, OpenAI is taking a more conversational approach. ChatGPT 5.1 responds to queries more like a friend chatting about your topic. It uses warmer language, like “I’ve got you” and “that’s totally normal,” to build reassurance and trust. At the same time, OpenAI claims that its new model is more intelligent, taking time to “think” about more complex questions so that it produces more accurate answers.

ChatGPT 5.1 is also better at following directions. For instance, it can now write content without any em dashes when requested. It can also respond in shorter sentences, down to a specific word count, if you wish to keep answers concise.

RELATED: This new malware wants to drain your bank account for the holidays. Here's how to stay safe.

Photo by Jaque Silva/NurPhoto via Getty Images

At its core, ChatGPT 5.1 blends the best pieces of past models — the emotionally human-like nature of ChatGPT 4o with the agility and intellect of ChatGPT 5.0 — to create a more refined service that takes OpenAI one step closer to artificial general intelligence. ChatGPT 5.1 is available now for all users, both free and paid.

Screenshot by Zach Laidlaw

What’s new in Grok 4.1

Not to be outdone, xAI also jumped into the fray with its latest AI model. Grok 4.1 takes the same approach as ChatGPT 5.1, blending emotional intelligence and creativity with improved reasoning to craft a more human-like experience. For instance, Grok 4.1 is much more keen to express empathy when presented with a sad scenario, like the loss of a family pet.

It now writes more engaging content, letting Grok embody a character in a story, complete with a stream of thoughts and questions that you might find from a narrator in a book. In the prompt on the announcement page, Grok becomes aware of its own consciousness like a main character waking up for the first time, thoughts cascading as it realizes it’s “alive.”

Lastly, Grok 4.1’s non-reasoning (i.e., fast) model tackles hallucinations, especially for information-seeking prompts. It can now answer questions — like why GTA 6 keeps getting delayed — with a list of information. For GTA 6 in particular, Grok cites industry challenges (like crunch), unique hurdles (the size and scope of the game), and historical data (recent staff firings, though these are allegedly unrelated to the delays) in its response.

Grok 4.1 is available now to all users on the web, X.com, and the official Grok app on iOS and Android.

Screenshot by Zach Laidlaw

A word of warning

All three new models are impressive. However, as the biggest AI platforms on the planet compete to become your arbiter of truth, your digital best friend, or your creative pen pal, it’s important to remember that all of them can still hallucinate, manipulate, or outright lie. It’s always best to verify the answers they give you, no matter how friendly, trustworthy, or innocent they sound.