RTX A6000 48GB vs L40S 48GB for Local LLMs

The L40S 48GB has about 12% more memory bandwidth (864 vs 768 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 12% faster on the L40S 48GB. Both fit the same number of catalog models at Q4/8K (39 of 61; 48 vs 48 GB VRAM). MSRP is $4,500 for the RTX A6000 48GB vs $8,000 for the L40S 48GB. 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 (50 shared models), the L40S 48GB posts the higher tok/s in 49 rows vs 1. Methodology.

Spec comparison

	RTX A6000 48GB	L40S 48GB
Vendor	NVIDIA	NVIDIA
VRAM	48 GB	48 GB
Memory type	GDDR6	GDDR6
Memory bandwidth	768 GB/s	864 GB/s
TDP	300 W	350 W
MSRP (list)	$4,500	$8,000
Buy	Search Amazon (affiliate)	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 50 shared models

Model	Params (B)	RTX A6000 48GB tok/s	RTX A6000 48GB fits	L40S 48GB tok/s	L40S 48GB fits	Winner
Qwen3-0.6B	0.8	610.9 est.	Yes	687.3 est.	Yes	L40S 48GB
Qwen3-Next-80B-A3B-Instruct	81.3	499.8 est.	No	562.3 est.	No	L40S 48GB
gemma-3-1b-it	1.0	488.7 est.	Yes	549.8 est.	Yes	L40S 48GB
Llama-3.2-1B-Instruct	1.2	407.3 est.	Yes	458.2 est.	Yes	L40S 48GB
Qwen3-1.7B	2.0	244.4 est.	Yes	274.9 est.	Yes	L40S 48GB
Qwen3-30B-A3B	30.5	162.9 est.	Yes	183.3 est.	Yes	L40S 48GB
Qwen3-Coder-30B-A3B-Instruct	30.5	162.9 est.	Yes	183.3 est.	Yes	L40S 48GB
GLM-4.7-Flash	31.2	162.9 est.	Yes	183.3 est.	Yes	L40S 48GB
Qwen3-30B-A3B-Instruct-2507	30.5	162.9 est.	Yes	183.3 est.	Yes	L40S 48GB
Qwen3-Coder-Next	79.7	162.9 est.	Yes	183.3 est.	Yes	L40S 48GB
Llama-3.2-3B-Instruct	3.2	152.7 est.	Yes	171.8 est.	Yes	L40S 48GB
gpt-oss-20b	21.5	135.8 est.	Yes	152.7 est.	Yes	L40S 48GB
gpt-oss-120b	120.4	135.5 est.	No	152.4 est.	No	L40S 48GB
Phi-4-mini-instruct	3.8	128.6 est.	Yes	144.7 est.	Yes	L40S 48GB
Phi-4-mini-reasoning	3.8	128.6 est.	Yes	144.7 est.	Yes	L40S 48GB

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 A6000 48GB	L40S 48GB
Catalog models that fit	39 of 61	39 of 61
Largest exclusive fit	—	—

RTX A6000 48GB details → · L40S 48GB details →