GPT-4 vs Claude vs Gemini for coding — honest breakdown after 3 months of daily use

About this article

I am a solo developer who has been using all three seriously. Here is what I actually think: GPT-4o — Strengths: Large context window, strong at boilerplate, excellent JSON output. Function calling is rock solid. Weaknesses: Sometimes confidently wrong on obscure APIs. Claude 3.5 Sonnet — Strengths: Best at understanding existing code structure. When I paste a whole module and ask it to refactor, it gets the intent right more often. Better at explaining why it made a change. Weaknesses: Can b...