coder-v2

on RTX 3090

189

tok/sdecode

RTX 3090 (24GB) + AMD EPYC 7702P 64-Core Processor (64c) + 252GB

rank in tier

1/4RTX 3090 runs

best workload

chat-short

where the rig flew

slowest workload

chat-long77.6 tok/s

where the rig struggled

backend

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

faster than

DeepSeek-Coder-V2-Lite-Instruct-4bit on M3 Ultra168 tok/s
qwen2.5-coder on RTX 4090161 tok/s
llama3.1 on RTX 4090154 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.