Skip to content

Can DGX Spark 64G run K2 Horizon 32B?

Comfortable: GGUF Q8_0 needs 39.7 GB at 4K, batch 1; headroom 24.3 GB.

Text input, FP16 KV cache, one card. Unified-memory capacity uses the GPU guide’s usable limit; negative headroom may require host offload. Speed source settings may differ from this 4K estimate.

Measured runs

No measured run for this pair. Sizing is an estimate, not a runtime test.

Quant memory at 4K

QuantTotal GBVerdictWeight sizing
GGUF Q4_K_M23.1ComfortableVerified file
GGUF Q5_K_M26.9ComfortableVerified file
GGUF Q6_K30.9ComfortableVerified file
GGUF Q8_039.7ComfortableVerified file

Context capacity · GGUF Q8_0

ContextTotal GBVerdict
4K39.7Comfortable
16K43.0Comfortable
32K47.4Comfortable
128K73.8Does not fit

Generation speed

GGUF Q8_0: 8 tok/s · Bandwidth ceiling (not measured)

Other model / GPU pairs