Where’s the logic in this at all? If a simple word like “Hi” takes that many tokens—are you guys actually developers, or do you just have nothing better to do?
This situation is so absurd it’s ridiculous.
I mean it was probably a one time error 
Was “hi” your first turn of input?
The error message gives you a hint of what’s going on.
You likely sent a parameter for max_output_tokens - the maximum bill for an API call that you want before the output is truncated.
The rate limiter treats that as the API call’s potential use of tokens.
The token-per-minute rate limit you are up against is exceptionally low: 100 tokens per minute. Did you set that limit yourself on the project? Do you not yet have a funded API organization, by an initial purchase of credits, that has graduated to “tier 1”?
| Tier | RPM | TPM | Batch queue limit |
|---|---|---|---|
| Free | Not supported | ||
| Tier 1 | 500 | 500,000 | 5,000,000 |
| Tier 2 | 5,000 | 2,000,000 | 20,000,000 |
You might be able to say “hi” if you don’t use any max_tokens type of limit in your API request. Some new accounts that have gone through the platform site onboarding can get some minimum testing calls - the exact nature of which is not documented, but usually using a mini model with the API call that is suggested you make during the “start building” presentation.