[BUG] 400 Error with Opencode Go & gpt-6-luna #1979

Problem (one or two sentences)

When using the Opencode Go subscription and selecting the GPT 6 Luna model, every request fails saying "Model does not support this protocol."

Context (who is affected and when)

Affects anyone attempting to use GPT 6 Luna with the Opencode Go provider.

Reproduction steps

1. Open Zoo Code, click Settings > Providers and add a new provider.
2. Configure API Provider as Opencode Go and enter a valid Opencode Go API Key.
3. Click refresh models.
4. Select Model gpt-6-luna.
5. Observe incorrect defaults: Context window: 200,000 tokens + "Does not support images" + "Does not support prompt Caching".
6. Start new session using the newly created model in Code Mode.
7. Submit a prompt, eg "create a readme for this project".

Expected result

Should see thinking and it should create a readme file, reporting its actions

Actual result

400 error "Model does not support this protocol." and a link for diagnostic information

Variations tried (optional)

Reproduced on two separate machines with two separate Opencode Go subscriptions:

Opencode Go for work, Ubuntu 26.04 + MATE + VS Codium and Zoo Code.
Opencode Go for personal, Linux Mint 22.2 + MATE + VS Codium and Zoo Code.

App Version

3.86.0

API Provider (optional)

Not Applicable / Other

Model Used (optional)

gpt-6-luna

Zoo Code Task Links (optional)

Related to #1431

Relevant logs or errors (optional)