Comparing Qubic's quorum consensus to traditional BFT - is this actually Byzantine Fault Tolerant in the formal sense?

Qubic uses a quorum-based consensus where 451 of 676 Computors must agree for a transaction to finalize. They describe this as Byzantine Fault Tolerant. I want to understand whether that description is technically accurate.

Classical BFT requires that the number of faulty nodes f satisfies n >= 3f+1, where n is total nodes. With 676 total and a 451 threshold, that means 225 nodes could theoretically fail or act maliciously and the network still reaches consensus. 676 >= 3(225)+1 = 676. That is exactly the BFT threshold - the minimum viable configuration.

The question I have is whether this is meaningfully different from PBFT or other BFT implementations at this scale, and what the practical attack surface looks like at 676 nodes versus something like Ethereum's validator set.

I am asking because "BFT" gets used loosely in crypto and I want to understand whether Qubic's specific implementation is formally correct or just approximately true.