Intent-based multi-model routing: autonomous agents manage Sparks-hosted Qwen, GLM, DeepSeek

jasonkneen · x · 2026-09-06

Developer Jason Kneen shares his evolving multi-model setup: two Sparks machines running MoE-style combinations of Qwen 3.8, GLM 5.3 and DeepSeek V4 Flash, swappable and usable in parallel.

With a custom proxy and playground, he no longer picks models — he picks an INTENT, and the system routes automatically using metrics, ratings and public data. Agents now run autonomously on the Sparks, maintaining, improving, optimizing, archiving and replacing services and models.

Original post →

More from coding & agent

coding & agent channel →