GPT-4 API - Early Access (Text to SQL purpose)

[dordonne.thomas]

I was using GPT-3.5 turbo API for Text to SQL purposes and just passed it to GPT-4.

My first thought:

1. The model now really respects everything I ask.
2. I get fewer syntactically incorrect requests. When I get one, it has difficulty to correct the error (have to ask 2 or 3 times).
3. The wider context (8k) allows giving more examples so the output is more accurate and reliable.
4. It is slower than 3.5-turbo but not slower than Davinci-003 2 or 3 weeks ago

Have to check for the rate limit to deploy in production.

Great job !

[PaulBellow]

Nice.

We’re trying to fence all the new GPT-4 API posts (early observations) here to make it easier for everyone to keep up.

Feel free to join us in the thread. Thanks.

GPT-4 Early Access -- First Thoughts

I am blown away by the difference the token amount makes. I am currently using the 8k model, which is almost 2x the size of the gpt-3 model and it makes a huge difference. Token compression aside, the increase in token amount sees an exponential increase in the power of this api. I believe the token limit’s effect on the api is akin to the effect that memory had on the early computer. If you have any questions on how the new model performs, please ask and I will be happy to answer!