← AI Hashrate 中文

Gaudi 3 128GB vs V100 32GB SXM2 for Local LLMs

The Gaudi 3 128GB has about 264% more memory bandwidth (3,276 vs 900 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 264% faster on the Gaudi 3 128GB. The Gaudi 3 128GB fits more of the catalog at Q4/8K — 57 vs 43 of 73 models (128 vs 32 GB VRAM). Largest model that fits the Gaudi 3 128GB but not the V100 32GB SXM2: Step-3.5-Flash (196.8B). MSRP is $12,500 for the Gaudi 3 128GB vs $10,000 for the V100 32GB SXM2. MSRP is a launch list price, not live retail — Amazon prices move; click through for the current price. Not retail-shoppable: Gaudi 3 128GB, V100 32GB SXM2 (datacenter-class; cloud rental is the realistic way to use it). In the Q4/8K head-to-head below (57 shared models), the Gaudi 3 128GB posts the higher tok/s in 57 rows vs 0. Methodology.

Spec comparison

Gaudi 3 128GBV100 32GB SXM2
VendorIntelNVIDIA
VRAM128 GB32 GB
Memory typeHBM2eHBM2
Memory bandwidth3,276 GB/s900 GB/s
TDP600 W300 W
MSRP (list)$12,500$10,000
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 57 shared models

ModelParams (B)Gaudi 3 128GB tok/sGaudi 3 128GB fitsV100 32GB SXM2 tok/sV100 32GB SXM2 fitsWinner
MiniCPM5-1B1.083113.9 est.Yes895.6 est.YesGaudi 3 128GB
Qwen3.5-2B2.01681.5 est.Yes483.6 est.YesGaudi 3 128GB
DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx2.31462.2 est.Yes420.5 est.YesGaudi 3 128GB
Qwen3.6-35B-A3B36.01121.0 est.Yes322.4 est.YesGaudi 3 128GB
Qwen3.5-35B-A3B34.71121.0 est.Yes322.4 est.YesGaudi 3 128GB
GLM-4.7-Flash31.21121.0 est.Yes322.4 est.YesGaudi 3 128GB
Qwen3-30B-A3B-Instruct-250730.51121.0 est.Yes322.4 est.YesGaudi 3 128GB
Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF1630.01121.0 est.Yes322.4 est.YesGaudi 3 128GB
Qwen3-Coder-Next79.71121.0 est.Yes268.3 est.NoGaudi 3 128GB
Ornith-1.0-35B35.01121.0 est.Yes322.4 est.YesGaudi 3 128GB
Qwen3-Next-80B-A3B-Instruct81.31121.0 est.Yes258.2 est.NoGaudi 3 128GB
qwen35b-a3b-fable-sft-abliterated34.71121.0 est.Yes322.4 est.YesGaudi 3 128GB
Llama-3.2-3B-Instruct3.21050.9 est.Yes302.3 est.YesGaudi 3 128GB
gpt-oss-20b21.5934.2 est.Yes268.7 est.YesGaudi 3 128GB
gemma-4-26B-A4B-it25.2885.0 est.Yes254.5 est.YesGaudi 3 128GB

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 128GBV100 32GB SXM2
Catalog models that fit57 of 7343 of 73
Largest exclusive fitStep-3.5-Flash (196.8B)—

Gaudi 3 128GB details → · V100 32GB SXM2 details →