andy is a software engineer publishing open-source projects through the Andyyyy64 account. WhichLLM helps people choose local language models by combining hardware-fit estimates with available benchmark evidence.[1][2]
Start with WhichLLM →
Links
GitHub 
X · Andyyyy64 Contributions
WhichLLM detects GPU, CPU and memory resources, then ranks models using estimated memory, speed and available quality measurements. It accounts for quantization, offloading and differences between active and total parameters in mixture-of-experts models. Hardware simulation also lets a user compare possible configurations.[3]
Ideas
The tool labels the origin and confidence of benchmark evidence rather than treating every uploader’s claim equally. Its defaults can recommend tight memory fits or partial offloading; stricter filters reserve headroom and require full GPU residency. These are estimates and recommendations, so a user should check the intended model and workload on the actual machine.[3]
Tools, writing & talks
Tools
Start here
WhichLLMgithub.comA local-model recommendation CLI using hardware estimates, model metadata and graded benchmark evidence.
Read the conservative GPU-only example and the evidence labels first. The recommendation command needs Python tooling; downloading or running a suggested model has separate storage and hardware requirements.
Creator[1][3]Original source