Skip to content

Can L40S 48G run Clover-35B-A3B-Preview?

Comfortable: GGUF Q8_0 needs 38.6 GB at 4K, batch 1; headroom 9.4 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 Q3_K_M18.2ComfortableVerified file
GGUF Q4_020.9ComfortableVerified file
GGUF Q4_K_M23.4ComfortableVerified file
GGUF Q5_K_M28.3ComfortableVerified file
GGUF Q8_038.6ComfortableVerified file

Context capacity · GGUF Q8_0

ContextTotal GBVerdict
4K38.6Comfortable
16K38.9Comfortable
32K39.2Comfortable
128K41.3Comfortable

Generation speed

GGUF Q8_0: 285 tok/s · Bandwidth ceiling (not measured)

Other model / GPU pairs