Skip to content

Can RTX 4090 run Recursive Flint 40B-A2.5B?

Comfortable: GGUF Q3_K_M needs 20.6 GB at 4K, batch 1; headroom 3.4 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_M20.6ComfortableVerified file
GGUF Q4_K_M26.7Does not fitVerified file
GGUF Q8_044.5Does not fitVerified file

Context capacity · GGUF Q3_K_M

ContextTotal GBVerdict
4K20.6Comfortable
16K21.0Comfortable
32K21.4Tight
128K24.3Tight

Generation speed

GGUF Q3_K_M: 866 tok/s · Bandwidth ceiling (not measured)

Other model / GPU pairs