Skip to content

Can RTX 3070 Ti run TwIL 8B balanced?

Comfortable: GGUF Q4_K_M needs 6.3 GB at 4K, batch 1; headroom 1.7 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_M6.3ComfortableVerified file

Context capacity · GGUF Q4_K_M

ContextTotal GBVerdict
4K6.3Comfortable
16K8.3Tight
32K11.1Does not fit
128K27.6Does not fit

Generation speed

GGUF Q4_K_M: 120 tok/s · Bandwidth ceiling (not measured)

Other model / GPU pairs