News for hackers. Stuff that matters.

A comment on Qwen3: Think deeper, act faster. Parent comment

Re:Qwen3: Think deeper, act faster by bionhoward on Monday April 28, 2025 @09:25PM (#43827884)
Wouldn’t it be 1GB (billion bytes) per billion parameters when each parameter is 1 byte (FP8)?

Seems like 4 bit quantized models would use 1/2 the number of billions of parameters in bytes, because each parameter is half a byte, right?

[ Parent | on HN ]
Re:Qwen3: Think deeper, act faster by daemonologist on Monday April 28, 2025 @11:08PM (#43828366)
Yes, it's more a rule of thumb than napkin math I suppose. The difference allows space for the KV cache which scales with both model size and context length, plus other bits and bobs like multimodal encoders which aren't always counted into the nameplate model size.
[ Parent | on HN ]