Hey team — I'm getting intermittent failures where calls crash mid-conversation: Error: Model sequence failed - all attempts exhausted Root cause: The model chatgpt-4o-latest does not exist or I do not have access to it (404 model_not_found) Call ID with detailed logs: 019c71af-54a1-7bb2-acc4-424bb553268b Critical finding from logs: Lines 259-115013: Model correctly shows "gpt-4o" (call working, multiple successful exchanges) Lines 115132, 115240: Model suddenly changes to "chatgpt-4o-latest" Result: Call crashes after ~3 minutes of working conversation What's confusing: My assistant config explicitly specifies gpt-4o, and it WORKS for multiple turns. Then mid-call, something on VAPI's backend switches to chatgpt-4o-latest and crashes the session. Questions: Is VAPI internally rewriting gpt-4o → chatgpt-4o-latest intermittently? Why would this happen mid-call after successful exchanges? I switched from "daily" to "weekly" channel — could this be a bad build? Config snippet: provider: openai model: gpt-4o firstMessage: null firstMessageMode: assistant-speaks-first-with-model-generated-message Happy to share full logs or additional call IDs showing the pattern.