Comment by rvnx

And somehow, when reading the benchmarks, Gemini Pro seems to be a regression compared to PaLM 2-L (the current Bard) :|