Dictation reports “Could not download” after successful local model installation on Apple Silicon

Does this issue occur when all extensions are disabled?: Not tested yet Report Issue' dialog can assist with this. --> - VS Code Version: 1.144.0 (Stable) - Commit: 2c33b50b16c472efe58357d1d2a8a5414bc23fbf - OS Version: macOS 26.7.1 (build 25G241), Apple Silicon (arm64) ### Steps to Reproduce 1. Run “Chat: Install Dictation Model from Local Package…” and select the Nemotron model package. 2. Observe the success notification: “Installed nemotron-3.5-asr-streaming-0.6b version 3.” 3. Click the microphone in Copilot Chat. 4. Observe the download error below. 5. Reload the window or fully quit and reopen VS Code, then retry. The error persists. ### Actual Result “Could not download the nemotron-3.5-asr-streaming-0.6b speech-to-text model, which can happen on networks that block the model registry. You can install it from a downloaded package instead.” ### Expected Result Dictation uses the locally installed model, or reports the specific underlying error if initialization fails. ### Installation Checks - Installed product.json specifies Foundry Local runtime 2.0.1. - Runtime 2.0.1 exists with .complete-darwin-arm64 containing 2.0.1. - Inspected runtime native binaries are arm64. - Model metadata registers the generic CPU model, version 3. - The model directory contains the ONNX models and data files. - No dictation.model override was found in user settings. - File presence was checked; successful runtime/model loading has not been verified. Latest Insiders build: Not tested.