inference speed matters more than benchmark scores for local models

r/LocalLLaMA
AI Research

After testing a bunch of local models for actual coding tasks i've come to the