OpenAI Launches New Coding-Focused GPT-4.1 Models

[PowerButton]

Wait wait wait… OpenAI is deprecating (at least on the API models) the recently announced GPT-4.5 because they just released GPT-4.1 which is superior? Hmmm why not call it GPT-4.6 if it’s superior?

If you look at the dash as a negative like GPT’negative’4.1 I guess it could be getting higher 🤓🤣

 

[svish]

Good to know about this but ChatGPT has a very complicated naming system for the different models. A simple and easy to remember one would have been good.

 

[MLVC]

Too many OpenAI models, streamline that stuff already.

 

[chrono1081]

No thanks. I’m a skilled developer I don’t need that garbage. AI code on all platforms is garbage unless you’re doing the most basic cookie cutter tasks.

Not only is AIs math always wrong, but it loves to create unnecessary methods, call methods that don’t exist, and just write garbage code.

 

[ZeroInspo]

No thanks. I’m a skilled developer I don’t need that garbage. AI code on all platforms is garbage unless you’re doing the most basic cookie cutter tasks.

Not only is AIs math always wrong, but it loves to create unnecessary methods, call methods that don’t exist, and just write garbage code.

Not saying that you “need it” need it at the moment, but these things only came into the scene like 3 years ago. Think about where it might be in 3 more. If you don’t at least find ways to incorporate it into your workflow you’re likely to find yourself outperformed by your peers. If I can get 80% of your skill at 10% of the cost, guess what I’m doing? Or a different scenario maybe, what If a much less skilled (and also cheaper) developer can produce code at your skill level with higher volume using AI, what do you think happens to you?

Not too long ago I saw someone from Y combinator saying how something like 80%+ of the code their current class of companies is using has been AI generated. People don’t really care if your code is artisanal hand crafted code, all they care about is whether it does what it’s supposed to do at the lowest cost possible in the shortest amount of time.

 

[johnsawyercjs]

Why is this even a choice in the end-user facing product? The api, I get it, but in ChatGPT why can't the AI figure out the best model to use for any particular question?

OpenAI is working on that, and supposedly ChatGPT 5 will do that:

 

[PowerButton]

No thanks. I’m a skilled developer I don’t need that garbage. AI code on all platforms is garbage unless you’re doing the most basic cookie cutter tasks.

Not only is AIs math always wrong, but it loves to create unnecessary methods, call methods that don’t exist, and just write garbage code.

It’s not perfect at writing code from scratch, but it’s super helpful for understanding and debugging stuff you didn’t write — Saves a ton of time when you’re thrown into unfamiliar codebases.

 

[karnac]

View attachment 2502169

They haven't learned about internal and external product name yet I guess. It appears they name these versions based on branches, which is fine for internal use. But when you're ready to push it out to the world you give it an external "product" name that makes sense. This isn't revolutionary.