Folks, how good is genspark.ai’s model at preserving fidelity?

I asked several of them, and they all said they were Sonnet 4.5 themselves.

You can check it here: