I've been experimenting with both versions, 3.5 and 3.7, and I'm curious about which one performs better for coding tasks. I've had some mixed experiences and would love to hear how others have found each model. Can someone provide insights on where one might excel over the other?
5 Answers
From my experience, 3.7 does tend to generate code a bit more efficiently, especially when given a clear task. However, it also has this annoying habit of going off on tangents—like, I asked it to discuss something first, and it just starts churning out a ton of code. I've had to get used to monitoring it closely to avoid token wastage. I recently switched back to ChatGPT after noticing that 3.7 was introducing bugs that I had to spend a lot of time debugging. 3.5 feels much more straightforward for certain tasks, especially when the prompts are simple and direct.
I totally get what you're saying! I've noticed that 3.7 can be a bit excessive sometimes, throwing in extra details or coding elements that weren't even part of the original request.
For creative writing tasks, 3.7 definitely shines, but when it comes to coding, it seems to go off the rails more often. In my case, 3.5 has been more reliable. The outputs from 3.7 can be like trying to drink from a fire hose—lots of information but not always what you needed!
Exactly! For coding, I find 3.5's responses are more manageable and straightforward, especially when I'm looking for precise revisions.
One key advantage with 3.5 is its knack for summarizing or recalling large amounts of info without issue, while 3.7 occasionally has hallucination problems, even when instructed to reference previous parts of the conversation. That alone makes 3.5 stand out for coding-related tasks.
That’s such a good point! I’ve had to deal with 3.7 misinterpreting prompts, which is frustrating, especially when I need clarity.
I prefer 3.7 when I'm using it for larger projects since it seems to produce more effective results, but it’s important to keep an eye on it so it doesn’t go down the wrong path too easily. For smaller tasks, I still turn to 3.5 given its simplicity.
I totally agree! I like using 3.5 for quicker jobs where I just want the job done without any drama.
Honestly, I’ve found that 3.5 handles coding tasks better than 3.7, which sometimes gets too complicated with its responses. While 3.7 can be efficient if you give it careful instructions, it tends to overcomplicate things and generates unnecessary steps.
Absolutely! I feel like 3.7 often overanalyzes and wastes time on solutions that didn't need changing.
Right? I've had issues where it just complicates simple tasks—it’s better to stick with 3.5 for more straightforward coding.
Sounds about right! I usually add "Do not change or update the code" at the end of my prompts for 3.7, which helps a lot.