Comment by sinuhe69

Recent incident with the Rio 3.5 model clearly shows that many coding models are specifically trained/fine tuned for the benchmarks.