Qualcomm Cloud AI 100 Ultra LLM inference performance slide deck (April 2024)

Fields

id

sources/qualcomm-cloud-ai100-ultra-llm-performance-apr2024

URL

https://www.qualcomm.com/content/dam/qcomm-martech/dm-assets/documents/LLM-Performance_APR2024.pdf

Publisher

Qualcomm

Kind

benchmark

Published

2024-04-01

Accessed

2026-10-14

notes

April 2024 vendor deck. FP16 compute, MXFP6 weights; 16x AI 100 Ultra per server. Latency-mode batch=1 (avg token latency + server tok/s) and throughput-mode batch=64. Vendor claim, not independently measured.

Notes

April 2024 vendor deck. FP16 compute, MXFP6 weights; 16x AI 100 Ultra per server. Latency-mode batch=1 (avg token latency + server tok/s) and throughput-mode batch=64. Vendor claim, not independently measured.

Cross-references

This record references no other records.