Default profile base model + full terminal use resets

[ocean basin]

I have added a custom endpoint with BYOK and set my "Default" profile (settings --> Agents --> Profiles --> edit) to use:

• Base model: GLM5.1 (custom)
• fll terminal use: GLM5-1 (custom)

But it randomly resets non stop to:
base model: auto (cost -efficient)
full terminal use: auto

im on windows 11

@brave grail

[brave grail]

Hi @ocean basin. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:

---

This is a known bug in Warp. The root cause has been identified: the LLMPreferences::reconcile_disabled_model_preferences function validates saved profile models only against the server-provided model list, but custom/BYOK models are stored in a separate custom_llms collection. Because your custom model UUID isn't found in the server list, the reconciler incorrectly treats it as "unusable" and clears the base_model field to null, which causes the UI to fall back to auto (cost-efficient). [Default profile model resets]

In short:

• Model picker includes both server models + custom/BYOK models ✅
• Reconciliation validates only against server models ❌ → custom model gets cleared

This has been confirmed by multiple users and is reproducible consistently with custom/BYOK model selections. Built-in models (e.g. claude-4-5-haiku) are not affected and persist correctly. [Settings revert to defaults]

A fix has been proposed in PR #11655 (referenced in issue #11564). You can follow along there for updates. [Default profile model resets]

In the meantime, there is no reliable workaround mentioned in the knowledge sources — the reset can happen on restart, when opening a new agent tab, or even during an active agent session.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon: