Which AI to Choose? A Personal Test Drive of GPT-5, Gemini, and Grok

Intro: The Problem of Choice

Today, using AI is becoming a way of life; AI helps solve and handle many everyday and business tasks. The list of all possibilities can be expanded indefinitely with that same AI.

I believe many of you have already asked yourselves: which AI model should I use? If you think about it, there is too much information, and studying it is not much fun, plus new things are released too often. This pushes us to make our choice unconsciously and use whatever comes to hand.

I don't want to overload you with information; I will try to describe everything briefly, clearly, and informatively. 🎯

1. The Test Contenders

We will look at 3 main models:

Gemini from Google
GPT from OpenAI
Grok from xAI

(I won't write about Claude from Anthropic, as I haven't personally used it, plus registration from the RF is only available with a foreign phone number 🚫).

2. Methodology: Paid Subscriptions Only

Over the last month, I have tried subscriptions for each of the models: Gemini from Google ($20), GPT from OpenAI ($20), xAI ($30). 💳

Only with a subscription can you assess the model's potential and its performance. There are subscriptions starting from $200 for Pro tariffs; I haven't used them. On X (formerly Twitter), those who use AI professionally (mainly for programming) often write that the difference [between tariffs] is significant and well worth it. 📈

3. Performance: Who is TOP Right Now?

The leader in performance right now is considered to be GPT with the release of version 5. 🏆 Gemini hasn't released new versions in a long time; Gemini 3 is now announced (to be released soon). In general, before GPT-5 appeared, I liked Gemini 2.5 Pro more.

By the way, I am currently using the Gemini 2.5 Flash API for my sites and bots—I've tested a lot, and I like it the most. ⚡️

GPT-5 currently handles almost all tasks best. Grok from xAI didn't impress me at all; I only liked the image and animation creation. + Sometimes I tried it when other models hit a wall, but it didn't help.

4. Objectivity vs. Flattery

By the way, GPT-5 is a "flattering" model, even when you ask for direct feedback. It will always be on your side and won't adhere to the truth, which can be misleading. 🎭 Grok and Gemini are more objective in this regard.

5. Limit Comparison

Regarding limits 📊:

GPT-5 [on subscription]: 3,000 messages per week (currently the highest).
Gemini: It's not enough for me (up to 100 requests per day on the PRO version).
Grok: 100 prompts every 2 hours.

6. Personal Preferences and Context Problems

In general, for solving ordinary tasks, writing tests, and code within a simple AI chat, I like Gemini 2.5 Pro the most (for its response speed and quality). ❤️ Although GPT-5 and Grok have objectively better performance, this is just my opinion, based more on the structure of the information delivery.

With Gemini, chats get overloaded with context very quickly (I sometimes change 10-20 chats a day). 🤯 Once [a chat] gets loaded, I create a brief summary in it, copy it, delete the chat, and start a new one by loading the summary.

(When writing code, this is natural: large volumes, and the context loads instantly). By the way, Gemini has a limit of up to 30,000 lines of code; if you upload more, the chat will crash without starting.

GPT has fewer problems in this area (even though Gemini's context window is 1,000,000 [tokens], and GPT-5's is 200,000).

Conclusion: Verdict and Expectations

Overall, GPT-5 from OpenAI is TOP right now. 👑

Most of all, I am waiting for the Gemini 3 model and Gemini Code Assist (when they make it proper, like Codex). ⏳

I must mention Codex from OpenAI (an AI agent for coding that can read, modify, and execute code, as well as perform commands in a sandboxed environment)—it's something else for writing code! 🔥 I won't go into detail; it's a separate and big topic, and besides, [I assume] few people are interested in it.