Skip to content

Can RTX 3080 10G run Ministral 3 14B Instruct?

Tight: GGUF Q4_K_M needs 9.3 GB at 4K, batch 1; headroom 0.7 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_M9.3TightVerified file
GGUF Q5_K_M10.7Does not fitVerified file
GGUF Q8_015.7Does not fitVerified file

Context capacity · GGUF Q4_K_M

ContextTotal GBVerdict
4K9.3Tight
16K11.4Does not fit
32K14.1Does not fit
128K30.6Does not fit

Generation speed

GGUF Q4_K_M: 64–85 tok/s · Derived range

Other model / GPU pairs