Anyscale again! They built a FREE model comparator, where you can evaluate 3 open source LLMs simultaneously using the same prompt. They support llama2 7b, 13b and 70b among 8 others.

https://lnkd.in/eUgnYdpn

Question though if anyone from Anyscale reads that, how do you manage system prompts and user prompt? Do we actually have to write the whole prompt in the text area, with the and [INST] tokens? If yes how do you deal with different models having different prompt formats?