Arc Pro B70 32GB vs H200 141GB SXM5 for Local LLMs
The H200 141GB SXM5 has about 689% more memory bandwidth (4,800 vs 608 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 689% faster on the H200 141GB SXM5. The H200 141GB SXM5 fits more of the catalog at Q4/8K — 55 vs 39 of 69 models (141 vs 32 GB VRAM). Largest model that fits the H200 141GB SXM5 but not the Arc Pro B70 32GB: Qwen3-235B-A22B-Instruct-2507 (235.1B). MSRP is $999 for the Arc Pro B70 32GB 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: H200 141GB SXM5 (datacenter-class; cloud rental is the realistic way to use it). In the Q4/8K head-to-head below (53 shared models), the H200 141GB SXM5 posts the higher tok/s in 53 rows vs 0. Methodology.