News for hackers. Stuff that matters.
Re:Mistral 7B Paper on ArXiv by matteoraso on Wednesday October 11, 2023 @08:17PM (#37852055)
>$0.001 per request second

This pricing is probably more expensive than gpt-3.5-turbo 4k context. A large prompt for the API would be 1k tokens in and 1k tokens out, which comes to $0.0035 for OpenAI. Your website says to expect a request to take 4 seconds minimum, so that's $0.004. Given how light Mistral is, I think you'd have to cut your price by at least a factor of 10 for it to be reasonable.

There are no comments on this comment yet.