RTX 4090 D

Nvidia · 24 GB VRAM

2recorded runs

70.9best tok/s

2sourced

How it compares — 24 GB tier

Best verified tok/s per GPU in this VRAM tier — compare any two like-for-like.

Evidence: 2 sourced runs · 1 independent source · first recorded Sep 4, 2026 · last recorded Sep 8, 2026. What these labels mean.

Model	Quant	Peak VRAM	Headroom
Qwen3.8-27B	Q4_K_M	23.5 GB	0.5 GB
Qwen3.8-27B	IQ3_S (GSQ-RCO)	20.7 GB	3.3 GB

Peak process memory recorded by the submitter on this card — evidence of what fit, not a guarantee. Plan with the VRAM calculator.

GPU	Model	Quant	Backend	Ctx	tok/s	prompt tok/s	VRAM GB	Source
RTX 4090 D · 24 GB	Qwen3.8-27B	Q4_K_M	llama.cpp b358-7620399	2421	70.9 ✓	1467.0	23.5	src
RTX 4090 D · 24 GB	Qwen3.8-27B	IQ3_S (GSQ-RCO)	llama.cpp c56093 (AtomicBot turboquant fork)	262144	67.5 ✓	—	20.7	src