Thread

Tweet 1

Is anyone noticing a lot of empty responses from Claude? Might be a litellm bug because I'm actually using the OpenAI python sdk and a litellm proxy. But the fact that in the logs I see the counts makes me think it might be an anthropic issue.

---

Tweet 2

Also getting occasional 500s but they're pretty rare

---

Tweet 3

Small sample sizes and I haven't been able to test everything I'd like because of these issues. But it has performed better than gpt-4 in the testing I have done. If you're looking for more data points

---