This is my read as well - Chinese labs don’t really seem to believe in AGI and instead the focus is on optimising the transformer architecture so it can be actually useful for specific tasks, ideally at a low cost