Julien Chaumond uses Pi's Virtual Models to route queries between local and remote LLMs

Julien Chaumond says he uses Pi's Virtual Models to route queries between local and remote models.

In an X post dated November 10, 2026, he says the setup sends requests to Qwen 3.8 27B, running locally in Llama, or to Kimi K3 or GLM 5.3, hosted remotely. He says the choice depends on query complexity and his machine's current load, and that it removes manual model switching.

Chaumond credits Pi developers @badlogicgames and @mitsuhiko and the ggml team. The post is his own account of personal use and does not include benchmarks or performance measurements.