I've been working on a simple Streamlit app to create graphs using different AI models. Initially, I used o4, which worked well for my needs. After hitting some limits, I switched to Gemini 2.5, and that model also performed great until I encountered limits again. Then I tried DeepSeek, and while it started off strong, it quickly began to make mistakes in the code and couldn't resolve some of the issues I encountered. Eventually, I returned to Gemini 2.5 and found that it could do everything DeepSeek couldn't. Is there really a significant difference between these models, or did I just have a string of bad luck with DeepSeek?
3 Answers
I think Gemini 2.5 is currently one of the top models out there. It really has a strong output, especially for coding tasks.
Honestly, I tried building a graph editor with DeepSeek too, and it started to go sideways pretty fast. I can't even imagine how people manage to create complex projects like Minecraft with it!
Gemini 2.5 Pro is definitely worth considering. Just remember, every AI has its quirks and limitations, so it's not all sunshine and rainbows with any of them.
I've seen it churn out massive code files effortlessly, which is super impressive!