Will Switching Models Make an Agent Forget?

Summary

Switching models does not automatically make Hermes Agent forget you because the model, persona, and memory are separate assets. The model decides how to reason. SOUL.md defines role and expression boundaries. USER.md and MEMORY.md retain the user profile and cross-session facts. The separate item to inspect is an automation that has pinned an older model.

If the answer sounds different, did memory disappear?

After a model switch, wording, speed, and tool choices can change noticeably. That can feel like "it no longer knows me". Changed behavior and lost persistent data are different things.

1. Three layers answer three different questions

Layer	What it owns	State after a model switch
Model	Reasoning and generation	Replaced
Persona	Role, tone, and boundaries	Preserved
Memory	Preferences, facts, shared history	Preserved

The new model interprets the same context differently, so the experience can change. It does not clear those independent files merely by being selected.

2. Why it can still feel different

Models differ in instruction following, context handling, tool use, and writing habits. If a new model is briefer or cites less background, it can appear forgetful. Confirm that Memory and Persona are intact before deciding whether the new model fits the task.

3. A post-switch checklist

Test a known preference in normal chat.
Confirm that the persona still behaves as intended.
Review unfinished automations for an old pinned model and decide whether to synchronize them.

Conclusion

The model is a replaceable reasoning engine. Persona and memory are the persistent layers that keep the agent recognizably yours.

Next: Where Hermes Agent's personality comes from.