Yes, I’ve handled this before. That error usually means the model instance isn’t available at that moment (capacity spike, routing issue, or temporary provider throttle), even if you're using GPT-5.2 Instant. Since it happens intermittently, it’s typically a scaling or failover configuration issue rather than a logic error in your flow. I can help you set up proper fallback models and retry handling so calls don’t fail when this happens. Feel free to DM me with your current pipeline setup. @Glenn_ZP