idk but naive me sees the article say the model needed 158GB of RAM so I guess 10x 16GB GPUs? Apparently retail price is $1k per 5080 16GB but they're actually selling at $1.5k so I guess $15k for just the GPUs.

▲ petu an hour ago | parent [-]

158GB number is for Mistral Medium 3.5 128B model, Muse Spark is way larger.

  Muse Spark Open would require more than 1.6TB of RAM in its native precision (BF16) for the model parameters alone. In a 4-bit quantized (NVFP4) format, the model memory footprint is lowered to about 493GB.

▲

bigyabai an hour ago | parent [-]

So about 24gb/expert at NVFP4. If you compressed the KV cache, you might be able to get away with 2x 24GB cards and stream the experts.

▲

petu an hour ago | parent [-]

If you're happy with generation speed in 5 t/s range, sure -- PCIe 5.0 x16 is only 64GB/s.

	▲	bigyabai 44 minutes ago \| parent [-]
		I never said it would be fast! In any case, I'd much rather run GLM 5.3 Flash on that sort of hardware. Better fit for the performance profile, and will probably beat Muse Spark Open in most uses.