Will test out the new model. Naming of these models could have been much better.
So it is at least as good as most politicians. Good to know.I’m so surprised by the number of people who are happy with ChatGPT. These past few weeks, I don’t think I’ve had a single correct answer from it, even for the simplest questions. It doesn’t take my settings into account, it hallucinates answers, it makes completely false statements, and it’s been totally unusable to the point that I cancelled my subscription. I had to ask it to check its answers every time, and every time I would get “You’re right, I made a mistake,” literally every single time. Even the translations were completely off (I’m French, I understand English perfectly but I’m not very good with grammar). Once it even translated into German....
Ah, but which model did you ask? because maybe 4o-mini is "great at recommending another model."I literally asked it which model to use for my needs and recommended the pros and cons... So I feel you can just do this to answer your comments (spoiler alert: 4.1 was the best and more accurate, especially with the newly expanded token limit).
I could confirm that GPT-4.1 can analyse images very well, and it can start the process of generating images. ChatGPT actually uses a different model named "GPT Image 1" to generate images.I'm not sure why OpenAI would switch all users to the 4.1 model, much less the mini version, since the "o" in "4o" stands for "omni", meaning it can take not just text as input, but also images and audio (but not video yet), and it can also include images in its responses (the other two can't), and it has a more conversational style. But maybe it's that more conversational style that OpenAI wants to get away from, or just keep it as a non-default option, after the debacle with the 4o model's flamboyantly unctuous responses that ended about a week ago when OpenAI "adjusted" that approach when people couldn't stand it.
I have this version
Feature | GPT-4 | GPT-4-turbo |
Speed | Slower | Faster 🚀 |
Cost (to use via API) | More expensive | Cheaper 💸 |
Efficiency | Standard | More optimized ⚙️ |
Context length | Up to 8K tokens (some 32K) | Up to 128K tokens 🧠 |
Availability in ChatGPT | Not used | Used in ChatGPT (pro) ✅ |
I’m so surprised by the number of people who are happy with ChatGPT. These past few weeks, I don’t think I’ve had a single correct answer from it, even for the simplest questions.
Interesting. I asked 4o if its 4.1 models could analyze images sent to it by the user, and generate images too, and it said no to both. Though I'm not surprised--I've asked ChatGPT various questions about itself over the past year or so, and sometimes it gets its replies wrong.I could confirm that GPT-4.1 can analyse images very well, and it can start the process of generating images. ChatGPT actually uses a different model named "GPT Image 1" to generate images.
Fr?In my experience, Gemini is much better than ChatGPT.
Seriously. It feels like they need to do a hard reset on the naming/versioning. I have a plus subscription, and it's still defaulting to 4o, with o3, o4-mini, and o4-mini-high as the main alternate choices. You have to go under "More Models" to find 4.1, 4.1-mini, and 4.5. I would love to assume that 4.5 is the latest greatest new hotness, but honestly I wouldn't put $100 on that bet. Is 4o better than o3? is o less than 1? or is it 15, since it's the letter "o"? why would I choose 4.1-mini over 4.1? Why would I use anything except 4.5? Does 4.5 have the same "advanced reasoning" as o3? 4.1 is for "quick coding", but o4-mini-high is "great at coding"... so I need to choose between quick and great code? Is there a model that does great quick code?
And biggest question of all, does any of this matter? If the latest version is more or less all the improvements of the previous versions all wrapped up in one, then why give all this choice? Conversely, if all these choices are all meaningful, then why is there virtually no guidance on when to use which one?
I have, and it's answers are all over the place and change. Once I asked 4.1, and it told me that 4o was the latest... all the while 4.5 is staring at me from the drop down.Ask these questions to ChatGPT, it will be happy to explain the advantages of each model.
It's weird. The mac app always defaults to 4o. The iphone app used to keep 4.1 selected for days or weeks, but I've noticed recently that it seems to revert back to 4o much more readily. I haven't used the web version for a long time, so I'm not sure there.Nearly a month after this article and I’m still waiting for 4.1 to be default… 🤣 I have to change it every time.