RTX 4090 24GB vs RTX 5070 Ti 16GB for Local LLMs

The RTX 4090 24GB has about 12% more memory bandwidth (1,008 vs 896 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 12% faster on the RTX 4090 24GB. The RTX 4090 24GB fits more of the catalog at Q4/8K — 33 vs 25 of 61 models (24 vs 16 GB VRAM). Largest model that fits the RTX 4090 24GB but not the RTX 5070 Ti 16GB: GLM-4.7-Flash (31.2B). MSRP is $1,599 for the RTX 4090 24GB vs $749 for the RTX 5070 Ti 16GB. MSRP is a launch list price, not live retail — Amazon prices move; click through for the current price. In the Q4/8K head-to-head below (44 shared models), the RTX 4090 24GB posts the higher tok/s in 36 rows vs 8. Methodology.

Spec comparison

	RTX 4090 24GB	RTX 5070 Ti 16GB
Vendor	NVIDIA	NVIDIA
VRAM	24 GB	16 GB
Memory type	GDDR6X	GDDR7
Memory bandwidth	1,008 GB/s	896 GB/s
TDP	450 W	300 W
MSRP (list)	$1,599	$749
Buy	GIGABYTE GeForce RTX 4090 Gaming OC 24GB (affiliate) · More listings	Search Amazon (affiliate)

MSRP is the launch list price, not live retail. Buy links are Amazon affiliate links — the price after you click is the live price.

Q4 / 8K head-to-head — top 15 of 44 shared models

Model	Params (B)	RTX 4090 24GB tok/s	RTX 4090 24GB fits	RTX 5070 Ti 16GB tok/s	RTX 5070 Ti 16GB fits	Winner
Qwen3-0.6B	0.8	801.8 est.	Yes	712.7 est.	Yes	RTX 4090 24GB
gemma-3-1b-it	1.0	641.5 est.	Yes	570.2 est.	Yes	RTX 4090 24GB
Llama-3.2-1B-Instruct	1.2	534.5 est.	Yes	475.2 est.	Yes	RTX 4090 24GB
Qwen3-30B-A3B	30.5	213.8 est.	Yes	411.5 est.	No	RTX 5070 Ti 16GB
Qwen3-Coder-30B-A3B-Instruct	30.5	213.8 est.	Yes	411.5 est.	No	RTX 5070 Ti 16GB
Qwen3-30B-A3B-Instruct-2507	30.5	213.8 est.	Yes	411.5 est.	No	RTX 5070 Ti 16GB
GLM-4.7-Flash	31.2	213.8 est.	Yes	395.0 est.	No	RTX 5070 Ti 16GB
Qwen3-1.7B	2.0	320.7 est.	Yes	285.1 est.	Yes	RTX 4090 24GB
gpt-oss-20b	21.5	225.2 measured	Yes	189.5 measured	Yes	RTX 4090 24GB
Llama-3.2-3B-Instruct	3.2	200.5 est.	Yes	178.2 est.	Yes	RTX 4090 24GB
Mistral-7B-Instruct-v0.3	7.2	183.3 measured	Yes	160.2 measured	Yes	RTX 4090 24GB
Qwen3-Coder-Next	79.7	170.4 est.	No	67.3 est.	No	RTX 4090 24GB
Phi-4-mini-instruct	3.8	168.8 est.	Yes	150.0 est.	Yes	RTX 4090 24GB
Phi-4-mini-reasoning	3.8	168.8 est.	Yes	150.0 est.	Yes	RTX 4090 24GB
gemma-3-4b-it	4.3	164.5 est.	Yes	146.2 est.	Yes	RTX 4090 24GB

est. = bandwidth-based estimate; measured = curated real benchmark that overrides the estimate. Fits = weights + KV(8K) + 1 GB overhead ≤ 95% of VRAM.

Fit coverage at Q4 / 8K

	RTX 4090 24GB	RTX 5070 Ti 16GB
Catalog models that fit	33 of 61	25 of 61
Largest exclusive fit	GLM-4.7-Flash (31.2B)	—

RTX 4090 24GB details → · RTX 5070 Ti 16GB details →