Comment by qsbuilder

same experience here. Benchmarks don't capture the magic of having a 24/7 native-speaking conversational partner in a rare language. That alone makes these models worth it