It seems plausible that it's a similar size model and that the 3x drop is just additional hardware efficiency/lowered margin.
Or just pressure from Gemini 3
Maybe it's AWS Inferentia instead of NVidia GPUs :)
Or just pressure from Gemini 3