Skip to content

Can L40S 48G run Ministral 3 14B Instruct?

Comfortable: GGUF Q8_0 needs 15.7 GB at 4K, batch 1; headroom 32.3 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_M9.3ComfortableVerified file
GGUF Q5_K_M10.7ComfortableVerified file
GGUF Q8_015.7ComfortableVerified file

Context capacity · GGUF Q8_0

ContextTotal GBVerdict
4K15.7Comfortable
16K17.8Comfortable
32K20.5Comfortable
128K37.0Comfortable

Generation speed

GGUF Q8_0: 63 tok/s · Bandwidth ceiling (not measured)

Other model / GPU pairs