Model Routing

Magnitude routes work across a lineup of open models. Each task goes to the model best suited for it — so no token is wasted and you get the right balance of capability and cost on every turn.

​

How it works

​

Leader — GLM 5.2

A strong all-rounder that handles everyday work. The leader runs the conversation, holds your intent, and delegates to workers. It represents about 50% of total tokens.

​

Workers — Multiple models

Fast and cheap for routine work, capable for complex work. The leader delegates most implementation tasks to workers. They represent about 50% of total tokens. Workers use a mix of models depending on the role.

​

Workers

Each worker is a tuned bundle of system prompt, scoped context, dedicated toolset, and a model + reasoning level. The leader figures out who should do what — you don’t pick agents or route tasks. For simple questions and small fixes, the leader handles them directly. For larger work, it delegates to workers. When a worker finishes, the leader reviews the output and decides what’s next.

​

Note on model lineup

The models above reflect the current routing assignments. We continuously benchmark new models and update the routing as better options become available. Model names in the CLI may differ from the display names here.