Hello @Kernel, we experimented with Mixtral for one of the downstream task, but unfortunately, it didn't perform as well as LLAMA2. That's why we opted for LLAMA2. However, we appreciate your suggestion and will certainly consider trying Mixtral for the next version.
