3 ms·
I hate that gpt-3.5-turbo is so cheap that using systems like guardrails is a sane thing to do. I can almost always prompt davinci-003 without guardrails in a w
by gamegoblin 3y ago
I hate that gpt-3.5-turbo is so cheap that using systems like guardrails is a sane thing to do. I can almost always prompt davinci-003 without guardrails in a way to get my exact schema 1-shot, whereas guardrails + 3.5-turbo will often consume 2-4x more tokens, but that still makes it significantly cheaper.
- brigadier132 3y agoThe problem people are having is hitting the rate limits for chat gpt.
- selcuka 3y agoIs there a rate limit for the paid API [1] too? [1] https://platform.openai.com/docs/models/gpt-3-5 https://platform.openai.com/docs/models/gpt-3-5
- gamegoblin 3y agoYou will get rate limit errors on all the models if you push hard enough. I always wrap the API with some exponential backoff retries to work around this.