3 ms·

Does it support Claude?

• QueensGambit 2y ago
  Not yet. @promptrepo/score relies on token-level logprobs, which OpenAI exposes for their models. But, Anthropic’s Claude currently doesn’t expose token-level confidence (like logprobs) in their API. So, we can’t support Claude until they do. We’d love to add support if/when Claude exposes this capability.