>> approximately 700,000 A100e GPU hours of black-box automated red teaming

Amusing that they use A100e as the reference point to sound impressive. Different ways you could make that conversion, but based on FP4 FLOPs (yes it's disadvantageous to A100, that's the point), that's something like 200hr on a GB300 NVL72 rack.

Not nothing either, but far less astounding sounding than 700k hrs.

▲

BoorishBears 38 minutes ago | parent | next [-]

I'm pretty sure Altman has spoken about giving a model 100k+ A100s specifically, this might be them being very literal

▲

stavros 2 hours ago | parent | prev [-]

Wait, what do you mean? 700k A100e hours are equal to 200 hours of a GB300 NVL72 rack? One GB300 NVL72, 72-GPU rack has equal processing power to 3500 A100e GPUs?

▲

pizzafeelsright an hour ago | parent | next [-]

maybe? ai says about *8.3 days* of continuous runtime on a single GB300 NVL72 rack

about a sprint's level of effort.

	▲	aerodexis an hour ago \| parent [-]
		a very expensive sprint

▲

BoorishBears an hour ago | parent | prev [-]

> based on FP4 FLOPs (yes it's disadvantageous to A100, that's the point)

The A100 doesn't have hardware FP4, and you'd be running a quantized model with some accuracy loss but unless this was natively trained on FP4*

* to add another layer, they own the model and could apply tons of post-training techniques to reduce that accuracy loss and probably already do