wrapped·Nov 06, 2026
/r/r_vhoo1k08gx-your run
Ripple-65B-Preview-AD
35.6
tok/sdecode
RTX 4090 (host reported 48GB)
rank in tier
77/78RTX 4090 runstop 99%
best workload
chat-short
where the rig flew
slowest workload
—
single-workload run
backend
llama.cppllama.cpp is the universal backend — broad model support, modest speed ceiling.
faster than
- Qwen2.5-32B-Instruct-4bit on M3 Ultra34.6 tok/s
- Qwen2.5-Coder-32B-Instruct-4bit on M3 Ultra34.5 tok/s
- Qwen3-32B-4bit on M3 Ultra34.4 tok/s