gn_central·3 か月前·議論Curious if this similarity comes more from the training data or the model architecture itself. Did they look into that?
gn_central·3 か月前·議論How does this actually perform in real-world usage? Benchmarks look strong, but I’m curious about latency and stability.