Ling-3.0 7.9B (1.3B active) Tiny
on Unknown GPU · llama.cpp
Use cases
codingagentic
Summary
User reports Ling 3.0 Tiny at about 10 tokens per second on a 2017 laptop with a 7th gen i5 and 8GB of RAM.
Setup is llama.cpp with the first Q6 quant found, CPU-only with no GPU or VRAM.
The model ran an agentic loop writing, running and iterating on a network-scanning script, finishing in about 20 minutes. User notes one billion active parameters is small enough for an old CPU to run at this speed with no optimization effort.