← AI Hashrate

Gaudi 3 128GB vs H200 141GB SXM5 for Local LLMs

The H200 141GB SXM5 has about 47% more memory bandwidth (4,800 vs 3,276 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 47% faster on the H200 141GB SXM5. The H200 141GB SXM5 fits more of the catalog at Q4/8K — 48 vs 47 of 61 models (141 vs 128 GB VRAM). Largest model that fits the H200 141GB SXM5 but not the Gaudi 3 128GB: MiniMax-M2.7 (228.7B). MSRP is $12,500 for the Gaudi 3 128GB vs $35,000 for the H200 141GB SXM5. MSRP is a launch list price, not live retail — Amazon prices move; click through for the current price. Not retail-shoppable: Gaudi 3 128GB, H200 141GB SXM5 (datacenter-class; cloud rental is the realistic way to use it). In the Q4/8K head-to-head below (59 shared models), the H200 141GB SXM5 posts the higher tok/s in 58 rows vs 1. Methodology.

Spec comparison

Gaudi 3 128GBH200 141GB SXM5
VendorIntelNVIDIA
VRAM128 GB141 GB
Memory typeHBM2eHBM3e
Memory bandwidth3,276 GB/s4,800 GB/s
TDP600 W700 W
MSRP (list)$12,500$35,000
Buy——

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 59 shared models

ModelParams (B)Gaudi 3 128GB tok/sGaudi 3 128GB fitsH200 141GB SXM5 tok/sH200 141GB SXM5 fitsWinner
Qwen3-0.6B0.82605.9 est.Yes3818.2 est.YesH200 141GB SXM5
gemma-3-1b-it1.02084.7 est.Yes3054.5 est.YesH200 141GB SXM5
Llama-3.2-1B-Instruct1.21737.3 est.Yes2545.5 est.YesH200 141GB SXM5
Qwen3-1.7B2.01042.4 est.Yes1527.3 est.YesH200 141GB SXM5
Qwen3-30B-A3B30.5694.9 est.Yes1018.2 est.YesH200 141GB SXM5
Qwen3-Coder-30B-A3B-Instruct30.5694.9 est.Yes1018.2 est.YesH200 141GB SXM5
GLM-4.7-Flash31.2694.9 est.Yes1018.2 est.YesH200 141GB SXM5
Qwen3-30B-A3B-Instruct-250730.5694.9 est.Yes1018.2 est.YesH200 141GB SXM5
Qwen3-Next-80B-A3B-Instruct81.3694.9 est.Yes1018.2 est.YesH200 141GB SXM5
Qwen3-Coder-Next79.7694.9 est.Yes1018.2 est.YesH200 141GB SXM5
Llama-3.2-3B-Instruct3.2651.5 est.Yes954.5 est.YesH200 141GB SXM5
gpt-oss-20b21.5579.1 est.Yes848.5 est.YesH200 141GB SXM5
Phi-4-mini-instruct3.8548.6 est.Yes803.8 est.YesH200 141GB SXM5
Phi-4-mini-reasoning3.8548.6 est.Yes803.8 est.YesH200 141GB SXM5
gemma-3-4b-it4.3534.5 est.Yes783.2 est.YesH200 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

Gaudi 3 128GBH200 141GB SXM5
Catalog models that fit47 of 6148 of 61
Largest exclusive fit—MiniMax-M2.7 (228.7B)

Gaudi 3 128GB details → · H200 141GB SXM5 details →