deepseek-v4-flash-vision-exp succeeds even when "Enable models hosted in China" is disabled

Description

With "Enable models hosted in China" disabled, I selected opencode-go/deepseek-v4-flash-vision-exp in OpenCode CLI and an inference request completed successfully.
I do not know the upstream provider or inference execution region for this model.
Please clarify whether this behavior is intentional and document the model’s inference hosting region.
If inference for this model is executed in China, please require the China-hosting opt-in.

Related issues are #39845, #39872, and #39847.

Plugins

None

OpenCode version

1.18.23

Steps to reproduce

1. Subscribe to OpenCode Go.
2. Open the Go settings in OpenCode CLI.
3. Disable "Enable models hosted in China" and save the setting.
4. Re-open OpenCode and verify that the setting remains disabled.
5. Select opencode-go/deepseek-v4-flash-vision-exp.
6. Send a simple text-only prompt, such as Reply with exactly: test.
7. Confirm that the request completes successfully and returns a response.

Expected behavior

If inference for this model is executed in China, the request should be rejected while "Enable models hosted in China" is disabled.
If inference is executed outside China, OpenCode documentation should state the model’s actual inference hosting region.

Actual behavior

The request completed successfully while "Enable models hosted in China" was disabled.
The model’s upstream provider and inference execution region are not publicly confirmed.

Operating System

Ubuntu 22.04.5 LTS (x86_64)

Start with the OpenCode CLI Go settings and the model-selection path for opencode-go/deepseek-v4-flash-vision-exp. Confirm the model’s upstream provider and inference region, then determine whether the disabled China-hosting setting should reject the request or whether documentation should state the region. Done means the behavior matches the confirmed hosting location and is documented.