What's the best AI model for coding?
by•
New AI models pop up every week. Some developer tools like @Cursor, @Zed, and @Kilo Code let you choose between different models, while more opinionated products like @Amp and @Tonkotsu default to 1 model.
Curious what the community recommends for coding tasks? Any preferences?
3.3K views
Replies
I don't think there are big gaps between those models; it depends on the problems I am trying to solve. I usually start with Sonnet 4.5 and try others if it fails to resolve my problem. In cursor, they allow you to run multiple models in parallel, actually
@flyaway have you experimented with Composer 1? any feedback/thoughts?
@fmerian No, actually. But will try and keep u posted ;)
@flyaway looking forward to your thoughts!
I've tried different models over the course of one year, and I fell in love with anthropic, I feel it's the best model when it comes to complex coding and understanding.
@Claude by Anthropic is leading the way
I thought Claude was top contender, but I've had bugs all day there and it freezes on me. Maybe I'm just used to ChatGPT now.
Definetly opus 4.5, but also shotout to MiniMax M2, thing is Opus is just soo good, and claude code harness is just amazing to work with
@nicoeft never experimented with @Minimax M2 tbh. it's available for free on @Kilo Code -- might give it a spin. thanks!
I lean toward Sonnet 4.5 for correctness—it rarely messes up. Same time, Gemini sometimes surprises me with really smart suggestions.
I have tried every single one in the last month trying to vibecode my app....gemini 3 flash worked the best.
It's not that simple, but can mostly be summarized into two categories: Daily and Complex.
Gpt 5.2 is in my opinion the best model for Complex work and is freaking unstoppable combined with cursor ai debug mode. Why not Opus 4.5? well, it seems to struggle slightly with image reading and ui, which is important for complex tasks.
Claude 4.5 or even gemini 3 flash is the best daily model for daily stuff.
opus 4.5 for coding gemini 2 for animations and then chatgpt 5 for general use output logos -images so forth
Noodle Seed
From my experience Opus is really good at generating code for a specific framework or service, but for everyday tasks like refactoring, writing helpers, or SQL queries, GPT-5.2 ends up being faster and cheaper
Overall you pick the model for the task, not the other way around