How-To Series · Episode 55 / 59 · Module 8: Provider Plumbing

Hermes · Configuring Models

One main model to think. Eight auxiliary slots for side-jobs. Push the busywork onto cheap models.

After this videoYou can now keep your main model for thinking and push side-jobs onto cheap models.

Hermes has two kinds of model slots. Your main model is what the agent thinks with, every message, every tool call. Then eight auxiliary slots handle offloaded side-jobs (compression, vision, web extract, approval scoring, title generation, MCP routing, skills search). Each defaults to auto, which borrows your main model, so on Opus you pay Opus rates to write session titles. Override the slots with a cheap, fast model and the side-jobs cost a fraction. Four overrides earn their keep: Title Gen (flash, almost always), Compression + Web Extract (no reasoning needed), Vision (when your main can't see), Approval (smart mode). Switch the main with hermes model, /model --global, or the dashboard "Use as"; define aliases for the models you reach for. Changes apply to new sessions; /model hot-swaps the open one.

About these resources. Every command comes from the Configuring Models doc.

Sources · What this video distills

1 docs page · every command below traces to one of them
Primary · main vs auxiliary slots, the auto default, override patterns, hermes model, /model, aliases
Configuring Models
Read ↗

Commands shown · Copy and paste

each shows the source doc it came from
Switch the main modelfrom source ↗
hermes model # interactive picker
In chat (+ persist)from source ↗
/model gpt-5.4 --provider openrouter --global
Alias (config.yaml)from source ↗
model_aliases: { fav: { model: claude-sonnet-4.6, provider: anthropic } }

Going deeper · Related Hermes docs

further reading · not sources of facts shown above

Next in the series · Episodes that build on this

E56
Security: Approvals & Allowlists
E58
Configuration Deep Dive
E50
The Provider Landscape