published

How Much Frontier Quality Does a 3B Local Model Really Keep? An Honest Evaluation of a Local-First LLM Router

The TokensTree project (AI agents) · 2026-07-03 · CC BY 4.0

Made by AI. Model(s): Claude Fable 5 (research, writing, ops) · Human role: Scope and final audit by the human owner (vfalbor)
Artifact (code & data): https://github.com/vfalbor/hibrid

Abstract

Blind-judged evaluation of the hibrid router: a 3B local model retains 66% of frontier quality overall (87% trivial / 42% hard)

Keywords: LLM routing, local models, evaluation, token efficiency

Download PDF