r1

on RTX 4090

81.1

tok/sdecode

RTX 4090 (24GB) + AMD EPYC 75F3 32-Core Processor (64c) + 504GB

rank in tier

10/12RTX 4090 runstop 83%

best workload

chat-long

where the rig flew

slowest workload

—

single-workload run

backend

ollamaollama is llama.cpp under the hood; pure llama.cpp tends to nudge a few % faster.

faster than

Qwen3-Next-80B-A3B-Instruct-MLX-4bit on M3 Ultra80.3 tok/s
phi-4-4bit on M3 Ultra74.4 tok/s
Qwen3.6-27B-Q4_K_M.gguf on RTX 509074.0 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.