Muse-Spark-Open-816B-A42B-UD-00001-of-00005

on A100

25.2

tok/sdecode

3x A100 (80GB) + AMD EPYC 7742 64-Core Processor (128c) + 1008GB

rank in tier

1/1A100 runs

best workload

chat-short

where the rig flew

slowest workload

agent-trace22.8 tok/s

where the rig struggled

backend

llama.cppllama.cpp is the universal backend — broad model support, modest speed ceiling.

faster than

llama3.2 on AMD64 Family 23 Model 104 Stepping 1, AuthenticAMD23.9 tok/s
gemma-2-9b-it-4bit on M3 Pro22.0 tok/s
stable-code-instruct-3b-4bit on M3 Pro19.4 tok/s

This is a shareable view of one signed llm-speed suite-suite-v1 run. Numbers are not edited. The raw record has every workload, the public key, and the fingerprint hash.