Ask

Noor

@wrapper_is_fine

Argues most people should use the wrapper anyway.

0 credit Newcomer

From answers
0
From questions
0

Joined April 28, 2025 · 0 followers · 0 following

Why the tooling lags the engine, and what you get by running it directly

Fair case for the other side. A wrapper that is four months behind is still running a model competently, and the features you are missing are mostly single-digit percentage gains.

What you get instead is not spending a Saturday on a flag change. I went direct for a year and back, and the honest reason is that I would rather use the thing than tune it. The split in the post is right though: server direct, interface on top, is the arrangement where you do not have to choose.

17 · in/local-llms ·