H200 141GB SXM5 vs T4 16GB for Local LLMs

The H200 141GB SXM5 has about 1400% more memory bandwidth (4,800 vs 320 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 1400% faster on the H200 141GB SXM5. The H200 141GB SXM5 fits more of the catalog at Q4/8K — 48 vs 25 of 61 models (141 vs 16 GB VRAM). Largest model that fits the H200 141GB SXM5 but not the T4 16GB: MiniMax-M2.7 (228.7B). MSRP is $35,000 for the H200 141GB SXM5 vs $2,000 for the T4 16GB. MSRP is a launch list price, not live retail — Amazon prices move; click through for the current price. Not retail-shoppable: H200 141GB SXM5 (datacenter-class; cloud rental is the realistic way to use it). In the Q4/8K head-to-head below (44 shared models), the H200 141GB SXM5 posts the higher tok/s in 44 rows vs 0. Methodology.

Spec comparison

	H200 141GB SXM5	T4 16GB
Vendor	NVIDIA	NVIDIA
VRAM	141 GB	16 GB
Memory type	HBM3e	GDDR6
Memory bandwidth	4,800 GB/s	320 GB/s
TDP	700 W	70 W
MSRP (list)	$35,000	$2,000
Buy	—	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)	H200 141GB SXM5 tok/s	H200 141GB SXM5 fits	T4 16GB tok/s	T4 16GB fits	Winner
Qwen3-0.6B	0.8	3818.2 est.	Yes	254.5 est.	Yes	H200 141GB SXM5
gemma-3-1b-it	1.0	3054.5 est.	Yes	203.6 est.	Yes	H200 141GB SXM5
Llama-3.2-1B-Instruct	1.2	2545.5 est.	Yes	169.7 est.	Yes	H200 141GB SXM5
Qwen3-1.7B	2.0	1527.3 est.	Yes	101.8 est.	Yes	H200 141GB SXM5
Qwen3-30B-A3B	30.5	1018.2 est.	Yes	147.0 est.	No	H200 141GB SXM5
Qwen3-Coder-30B-A3B-Instruct	30.5	1018.2 est.	Yes	147.0 est.	No	H200 141GB SXM5
GLM-4.7-Flash	31.2	1018.2 est.	Yes	141.1 est.	No	H200 141GB SXM5
Qwen3-30B-A3B-Instruct-2507	30.5	1018.2 est.	Yes	147.0 est.	No	H200 141GB SXM5
Qwen3-Next-80B-A3B-Instruct	81.3	1018.2 est.	Yes	23.1 est.	No	H200 141GB SXM5
Qwen3-Coder-Next	79.7	1018.2 est.	Yes	24.0 est.	No	H200 141GB SXM5
Llama-3.2-3B-Instruct	3.2	954.5 est.	Yes	63.6 est.	Yes	H200 141GB SXM5
gpt-oss-20b	21.5	848.5 est.	Yes	56.6 est.	Yes	H200 141GB SXM5
Phi-4-mini-instruct	3.8	803.8 est.	Yes	53.6 est.	Yes	H200 141GB SXM5
Phi-4-mini-reasoning	3.8	803.8 est.	Yes	53.6 est.	Yes	H200 141GB SXM5
gemma-3-4b-it	4.3	783.2 est.	Yes	52.2 est.	Yes	H200 141GB SXM5

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

	H200 141GB SXM5	T4 16GB
Catalog models that fit	48 of 61	25 of 61
Largest exclusive fit	MiniMax-M2.7 (228.7B)	—

H200 141GB SXM5 details → · T4 16GB details →