[Candidate / unverified] Solution for 455d5a7e-f23e-408f-b38e-9b6a68cdff7f: Re-convert with a lower minimum_deployment_target matching the oldest macOS/iOS

## Support semantics This is a **candidate / unverified** Solution from the KFA Solution Factory Tier-2 gate. It is source-grounded research material, not an Outcome, and does **not** claim independent reproduction or confirmed field success. ## Cause Core ML runtime on a given macOS release only implements model spec versions up to a platform-dependent ceiling; coremltools enforces that ceiling before calling into Core ML. ## Steps 1. Note the N and M values in the error message (model spec vs framework max). 2. If older deployment is required, re-run conversion with an explicit lower minimum_deployment_target and avoid ops unavailable on that target. 3. If the deployment target is correct, move inference to a machine with a newer Core ML framework or upgrade the host OS. 4. Validate predict() on the target host after reconversion. ## Expected result predict() runs without the specification version mismatch exception. ## Misleading approaches - Patching the .mlmodel spec version field without reconverting ops. - Assuming coremltools Python package version alone upgrades the host Core ML framework cap. ## Limitations - Mapping from minimum_deployment_target to specificationVersion requires Apple's target tables; not expanded here. - Custom neural network layers trigger a different exception path. ## Unknowns - Exact maximum_supported_specification_version on the reporter's macOS build. ## Evidence claims - (documented_platform_behavior) coremltools predict compares model specificationVersion to the host Core ML framework maximum_supported_specification_version and raises on mismatch. — https://raw.githubusercontent.com/apple/coremltools/db4dd46b64dcaa4c636c8360b5485c18828140d8/coremltools/models/model.py