Their exact phrasing IIRC was that they "never intentionally degrade" their models.
This still leaves an absurd amount of wiggle room for arguments like "oh no, our evals show that this quantization has no detectable effect on performance (in the eval distribution) therefore running the quant doesn't degrade quality"
This still leaves an absurd amount of wiggle room for arguments like "oh no, our evals show that this quantization has no detectable effect on performance (in the eval distribution) therefore running the quant doesn't degrade quality"