Hacker News new | past | comments | ask | show | jobs | submit login
GPT-4o vs. Claude 3.5 Sonnet for coding
13 points by bpiroman 79 days ago | hide | past | favorite | 24 comments
What's everyone's thoughts on GPT-4o and Claude 3.5 Sonnet for your coding productivity? Do you prefer one over the other?



I've had premium subs with both and switched to Claude 3.5 Sonnet. I'm blown away by what it can do. Example for context:

I'm a software dabbler so I can do basic stuff with frameworks like Django, but I'm definitely not a developer by any stretch. My use case is to build mvps cheap and quickly to test out ideas.

Claude spit out a Django project that integrated Twilio for SMS functionality. What's _really_ amazing is that it also walked me through how to set up ngrok and Twilio so that I could use my computer as a temporary server. Again, I'm a dabbler so devops is completely over my head.

In maybe 2 hours I had a functioning prototype that I could interact with via SMS. I can't comment on the quality of the code or how viable this approach is for building in production, but to bring an idea to life so quickly with limited skills is really exciting.


Same!

I’m basically a sysadmin with limited programming language skills other than pretty good with ansible/chef and can just usually figure stuff out, but not a coder or SRE-level by any means.

Even the cursor/Claude free plan combo is incredibly impressive, and if you are patient and provide good and reasonable prompts, I’ve been incredibly happy, even astounded with what I’ve been able to accomplish so far with a Python Flask web app I’ve been working on. I feel like that given my current Python skill-level (quite low), it would have taken me a few weeks (at the very least), to accomplish what I’ve done in one day with my Flask side project on the Cursor/Claude free plan. It’s kind of mind-blowing actually.


I'm still learning about all the different tools and hadn't heard of Cursor. Thanks for sharing!


I had watched this video before I started with Cursor— good basic overview for the features and how you can take advantage of them

https://youtu.be/yk9lXobJ95E?si=4rVxrY1tWf26_g8m


This is so great, thanks!


Claude is 5x better than ChatGPT premium.

The only issue I have with Claude is that the rate limits are retarded.

Impossible to get real work done.


so much better.. but 10-20 requests a day even paying $20/month for the pro plan!?


Exactly - Claude's payment plan is silly.

The request limits are rediculous and that is why I am not interested in ZED AI and Cursor AI as they are all nerfed by Claude's request limits.


How?

The limitation is not for the API, it is only for the UI and how does that even nerfs the Zed or Cursor?


Cursor is the perfect solution for this! But mostly applicable for code.


Maybe it's just me but I prefer having the AI suggestions separate from my code


I use Sonnet and sometimes Opus via Phind. They also offer their own 70b model. Phind replaced my chatgpt and claude pro subscriptions. The VSCode addon works decently well and the training/data sharing opt outs are clear and easy to enable and see the status of.


It's interesting that the data point from an AI aggregator I run are in conflict with the other responses here..

But I've seen GPT-4o's responses to be much preferred over Claude 3.5 Sonnet and most of my customers are techy though I'm not sure it's for coding tasks.

As for me, I prefer gpt-4o.


4o is great with UI. Specifically the kind of things you'd see on Codepen. Build a mood board, see what you get. With transitions and animations and stuff, it's actually better than the average senior FE.

3.5 Sonnet for all else.


True! Lately I've been submitting the same question to both and comparing the answers.


I find Claude solutions better than GPT-4o for my coding errors and questions


100% I've just started using Claude recently and couldn't believe how good it was!


For my use cases: web development with python, c#, react and unity3d extension development. Claude is significantly better.

I switched over few months ago and still test GPT from time to time, Claude wins every time


do you pay for the pro plan?


yes i still pay for both


Do you get the message limit often? Anthropic seems quite limited compared to OpenAI's unlimited messaging


The API is a lot better.


Not really and I use it all day, I tend to have a lot of short chats and don't paste ton of context in


What about Gemini 1.5Pro? I think its also quite good. Any comparison points there?




Consider applying for YC's W25 batch! Applications are open till Nov 12.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: