We've been working hard on making the Neural endpoints even better. The response time for the Signature and Conversation can be reduced to 300 ms if the message_id was used previously. This means if you make a request Signature and then to Conversation, we cache the results, and the results from Conversation return faster.