Provider update prompt repeats when custom models are preserved #8504

What happened?

Qwen Code repeatedly shows the Built-in Provider Update prompt after a provider update has completed successfully when the provider configuration contains a user-added custom model.

This is reproducible with Qwen Code v0.21.4 and the Standard API Key provider when qwen3.8-max is retained as a custom model:

1. Start Qwen Code with an existing Standard API Key configuration containing qwen3.8-max.
2. Select Update all in the built-in provider update prompt.
3. Observe Standard API Key configuration updated successfully.
4. Restart Qwen Code.

The same update prompt appears again.

What did you expect to happen?

After a successful update, the provider metadata version should match the built-in provider template version, so the prompt should not reappear until the built-in template changes again. User-added custom models should remain configured.

Why does it happen?

Update detection hashes only the current built-in provider template. The update path preserves custom model IDs by passing both built-in and custom IDs into the install plan, while the persisted provider metadata version is derived from every model in that plan. The stored hash therefore differs from the built-in-only hash used on the next startup.

For the v0.21.4 China Standard template, the startup hash is 194969866583427ea6906e911d295648bca46f1e1b292c817f9545922832e37d; preserving qwen3.8-max causes the update to persist e5091eabcede11affa65908c5b35f6d93a055bd68e416302b488caa677000a19 instead.

Suggested acceptance criteria

• A provider update records the version of the built-in provider template, independent of preserved custom models.
• Custom models remain configured after the update.
• Restarting with an unchanged built-in template does not show the update prompt again.

Environment

• Qwen Code v0.21.4
• Standard API Key / ModelStudio Standard
• Linux server