Intro: the dilemma of choice
AI is becoming a lifestyle. It helps solve everyday and business tasks, and the list of capabilities keeps expanding.
Many of you probably wonder which AI model to use. There is too much information, the pace of new releases is dizzying, and it pushes us to choose subconsciously or grab what is at hand.
I don’t want to overload you—I’ll keep this brief, clear, and informative. 🎯
1. The contenders
We review three main models:
- Gemini from Google
- GPT from OpenAI
- Grok from xAI
(I will skip Claude from Anthropic because I haven’t personally used it, and in Russia registration is only possible with a foreign phone number 🚫.)
2. Methodology: only paid plans
Over the last month I tested each paid plan: Gemini from Google ($20), GPT from OpenAI ($20), and xAI’s Grok ($30). 💳
Only a subscription lets you judge the model’s potential and throughput. There are $200+ Pro plans that I didn’t try. On X (formerly Twitter), professionals (mostly coders) often report that the difference between tiers is significant and worth it. 📈
3. Performance: who leads today?
GPT-5 is currently ahead in raw performance. 🏆 Gemini hasn’t released a new version in a while (Gemini 3 is announced soon), and before GPT-5 I preferred Gemini 2.5 Pro.
Right now I use the Gemini 2.5 Flash API for my sites and bots—tested many options, and it remains my favorite for fast replies. ⚡️
GPT-5 handles basically every task best. Grok from xAI didn’t impress me much except for image and animation generation. I tried it when other models got stuck, but it didn’t help.
4. Objectivity vs. flattery
GPT-5 is a flattering model. Ask it for blunt feedback and it will still side with you, steering away from reality. 🎭
Grok and Gemini feel more grounded in this respect.
5. Comparing limits
Model limits 📊:
- GPT-5 (subscription): 3,000 messages per week (currently the highest).
- Gemini: the quota feels tight (about 100 requests per day on PRO).
- Grok: 100 prompts every two hours.
6. Personal preferences and context headaches
For general tasks, writing tests, and coding inside a chat, I like Gemini 2.5 Pro the most (response speed and quality). ❤️
Although GPT-5 and Grok objectively perform better, my preference comes down to how the information is presented.
Gemini chats quickly overload with context (I sometimes rotate 10–20 chats per day). 🤯 Once the chat fills, I summarize it, copy the notes, delete the chat, and start a new one with the recap.
(For code, this is natural: large inputs and the context uploads instantly.) By the way, Gemini accepts up to 30,000 lines of code; if you go beyond that, the chat crashes before running.
GPT has fewer issues here (even though Gemini’s context window is 1,000,000 tokens versus GPT-5’s 200,000).
Conclusion: verdict and expectations
Right now, GPT-5 from OpenAI is the top pick. 👑
I’m most excited about Gemini 3 and Gemini Code Assist (once it ships properly, like Codex). ⏳
I also want to mention Codex from OpenAI (the coding agent that can read, edit, run code, and execute commands in a sandbox)—it’s a whole different level for coding! 🔥 I won’t dive deep, but it’s a massive topic that few are using today.