RTX 3090

Nvidia · 24 GB VRAM

25recorded runs

709.5best tok/s

18sourced

How it compares — 24 GB tier

RTX 3090709.5 RX 7900 XTX550.2 RTX 4090141.2 RTX 3090 Ti97.0 RTX 4090 D70.9

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

Evidence: 18 sourced runs · 7 awaiting review · 2 independent sources · first recorded May 19, 2026 · last recorded Oct 5, 2026. What these labels mean.

What fits on 24 GB — measured

Model	Quant	Peak VRAM	Headroom
Qwen3.8-27B	W4A16	22.4 GB	1.6 GB
Ornstein3.6-27B-MTP-NSC-ACE-SABER	Q4_K_M	22.0 GB	2.0 GB
NVIDIA-Nemotron-3.5-Lightning-30B-A3B	NVFP4	21.4 GB	2.6 GB
Qwen3.8-27B	Q4_K_XL	18.9 GB	5.1 GB
Qwen3.8-27B	EXL3-3.5bpw	17.3 GB	6.7 GB
gemma-4-12B-it-qat	UD-Q4_K_XL	11.8 GB	12.2 GB
Qwen3.5-9B-MTP	Q6_K_XL	9.1 GB	14.9 GB
Qwen3.5-4B-MTP	Q6_K_XL	5.6 GB	18.4 GB

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

Price history

used$1,720▲ 63.8%3 snapshots

Weekly street-price snapshots — all snapshots.

All benchmarks Compare RTX 3090 →

GPU	Model	Quant	Backend	Ctx	tok/s	prompt tok/s	VRAM GB	Source
RTX 3090 · 24 GB	LFM2.5-1.2B-Instruct	Q4_K_M	llama.cpp 0.3.0-dev (build 1, commit d7bd3bf)	24576	709.5 ✓	11527.5	2.4	src
RTX 3090 · 24 GB	Qwen3.5-0.8B-MTP	Q6_K_XL	llama.cpp 0.3.0-dev (build 1, commit d7bd3bf)	24576	476.1 ✓	6693.2	1.6	src
RTX 3090 · 24 GB	NVIDIA-Nemotron-3.5-Lightning-30B-A3B	NVFP4	llama.cpp 030ebb558 + PR 26623 @ 000d8ecd4 + l…	8192	353.7 ✓	1951.8	21.4	src
RTX 3090 · 24 GB	LFM2.5-2.6B	Q4_K_M	llama.cpp 0.3.0-dev (build 1, commit d7bd3bf)	24576	305.6 ✓	7609.8	2.3	src
RTX 3090 · 24 GB	Qwen3.5-4B-MTP	Q6_K_XL	llama.cpp 0.3.0-dev (build 1, commit d7bd3bf)	24576	240.4 ✓	2167.6	5.6	src
RTX 3090 · 24 GB	Qwen3.8-27B	EXL3-3.5bpw	exllamav3 1.4.2 (Mia fork 63b32f001d7b2cfed3b3…	24576	164.6 ✓	586.9	17.3	src
RTX 3090 · 24 GB	Qwen3.5-9B-MTP	Q6_K_XL	llama.cpp 0.3.0-dev (build 1, commit d7bd3bf)	24576	164.3 ✓	1760.2	9.1	src
RTX 3090 · 24 GB	Qwen3.8-27B	EXL3 4.00bpw	exllamav2 v1.5.0	8192	152.4 ✓	—	—	src
RTX 3090 · 24 GB	Qwen3.8-27B	EXL3 4.00bpw	buun-llama-cpp 1d0f493c73817176f6953069b7c10211de95…	262144	149.2 ✓	—	—	src
RTX 3090 · 24 GB	Qwen3.8-27B	AWQ-INT4 W4A16	sglang 2e7c85da6	8000	136.8 ✓	1616.0	—	src
RTX 3090 · 24 GB	Qwen3.8-27B	Q4_K_XL	llama.cpp 0.3.0-dev (build 1, commit d7bd3bf)	24576	81.9 ✓	517.1	18.9	src
RTX 3090 · 24 GB	Qwen3.8-Flash-Next	EXL3 2.50bpw	exllamav2 v1.5.0-compatible build with MoE CPU…	262144	38.6 ✓	—	—	src
RTX 3090 · 24 GB	Ornstein3.6-27B-MTP-NSC-ACE-SABER	Q4_K_M	llama.cpp buun-llama-cpp (86-real, MSVC 19.31,…	262144	32.9 ✓	98.9	22.0	src
RTX 3090 · 24 GB	GLM-5.3-Flash	EXL3 3.05 bpw	exllamav2 v1.5.1	131072	28.1 ✓	—	—	src
RTX 3090 · 24 GB	Qwen3.8-27B	EXL3 4.00bpw	buun-llama-cpp 1d0f493c73817176f6953069b7c10211de95…	150000	27.8 ✓	525.3	—	src
RTX 3090 · 24 GB	DeepSeek-V4.1-Flash	EXL3 3.0 bpw	vllm 0.13 (lazymio/vllm-backport, DeepSee…	262144	26.0 ✓	—	—	src
RTX 3090 · 24 GB	Qwen3.8-27B	EXL3 4.00bpw	exllamav2 v1.5.0	150000	24.5 ✓	—	—	src
RTX 3090 · 24 GB	Qwen3.8-Flash-Next	EXL3 2.50bpw	exllamav2 v1.5.0-compatible build with MoE CPU…	175000	20.9 ✓	—	—	src
RTX 3090 · 24 GB	NVIDIA-Nemotron-3-Nano-4B	Q4_K_M	llama.cpp —	512	171.3 …	5608.8	0.0	src
RTX 3090 · 24 GB	gemma-4-12B-it-qat	UD-Q4_K_XL	llama.cpp —	512	126.5 …	1871.8	11.8	src
RTX 3090 · 24 GB	Qwen3.8-27B	W4A16	vllm —	65536	113.2 …	—	22.4	src
RTX 3090 · 24 GB	DeepSeek-V4-Flash-0731-ROCmFP3	Q4_K_M	llama.cpp —	512	84.8 …	3465.8	—	src
RTX 3090 · 24 GB	DeepSeek-V4-Pro-Qwen3.5-9B-MTP	Q8_0	vllm —	2048	43.4 …	107.1	—	src
RTX 3090 · 24 GB	Qwen3.6-27B-MTP	Q4_K_M	llama.cpp —	2048	41.2 …	1309.1	—	src
RTX 3090 · 24 GB	DeepSeek-V4-Flash-0731	IQ2_XSS	llama.cpp —	2048	19.4 …	209.4	—	src

Related reading: GPU buying guide · used GPUs · Value index · Street prices