exo
exo copied to clipboard
[BUG] Deleting instance then creating new instance launches stale model
Describe the bug
Deleting an instance, creating a new one, then sends a chat completion request with the old, stale instance.
To Reproduce
Steps to reproduce the behavior:
- Create instance of Qwen3-30B
- Delete instance
- Create instance of Llama 3.2 3B
- Chat (model selector shows Llama 3.2 3B selected)
- It sends a request to wrong model, Qwen3-30B
Expected behavior
It should automatically select the most recently launched model. The model selector already shows the correct model. But it should actually use that model when chatting.
Actual behavior
It uses the wrong model, the previously launched model.
Environment
- macOS Version: 26.2
- EXO Version: c1be5184b2787bb11bc16c72789d0afd7b6b5510
- Hardware: Any
- Interconnection: Any