Skip to content

Can DGX Spark 64G run Qwen3.6 35B-A3B?

Comfortable: GGUF Q8_0 needs 39.6 GB at 4K, batch 1; headroom 24.4 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_M23.4ComfortableVerified file
GGUF Q8_039.6ComfortableVerified file

Context capacity · GGUF Q8_0

ContextTotal GBVerdict
4K39.6Comfortable
16K39.8Comfortable
32K40.2Comfortable
128K42.2Comfortable

Generation speed

GGUF Q8_0: 90 tok/s · Bandwidth ceiling (not measured)

Other model / GPU pairs