OpenAI has released a new model and we have decided to migrate to it.
You might have seen the amazing demo videos in which they interact with the AI model on their phones – talking to it. And you might have heard about the Scarlett Johansson voice debacle. Did or did they not steal her voice?
But you might not have heard how amazing the new model is at generating text.
We have run the tests and it’s amazing. So amazing we decided to switch (almost) all of our generating to GPT-4o.
Here are the top five reasons why we love GPT-4o
- The quality in English is better than GPT-4.
- The quality in other languages is way better
- It’s almost twice as fast as GPT-4
- It’s half the price of GPT-4-turbo
- It uses fewer tokens for the same amount of words
1: The quality difference in English
This is of course very difficult to measure. We have generated lots of texts using different models and compared them. We think it’s an improvement. But then again, it’s hard to pinpoint the exact differences, but it feels better when looking at it.
OpenAI has some kind of chart on their website, showing some graphs that it’s better. But it’s really hard to interpret.
2: The quality in other languages is way better
This was a huge improvement, especially in languages we work with often (like Swedish). There are fewer direct translations from English and weird sentences. Way to go OpenAI!
3: It’s almost twice as fast as GPT-4
This was very easy for us to notice. We log every call we make to OpenAI so we can see the numbers. And this model is fast. A faster modem means fewer problems when generating lots and lots of text. So thank you OpenAI!
This also means that every OpenAI-based AI online service just got twice as fast thanks to this model.
4: It’s half the price of GPT-4-turbo
This is also amazing. Thanks to this all of our Limitless clients will pay half as much for each article they generate.
5: It uses fewer tokens for the same amount of words
I haven’t read anything about this in the press, everybody seems to be preoccupied with the voice thing. I’m not sure how they achieved this but they managed to change the token count for generating text.
Again, they have a graph on their website. This time it’s a bit easier to understand:
These changes mean that if you are generating Turkish your token count gets a 23% reduction. Which means 23% lower cost.
Thank you OpenAI!
If you are a Limitless customer you will start benefiting immediately from the reduced token counts and the reduced costs of Tokens.
If you want to try it out, head over to our AI text-generating platform. It’s free to try!
