[API Gateway] Model `gpt-5.6-luna` returns HTTP 500 Internal server error via /zen/go/v1 endpoint

Description

Description

When calling the OpenAI-compatible API endpoint (https://opencode.ai/zen/go/v1/chat/completions) for model gpt-5.6-luna, the gateway immediately responds with HTTP 500 Internal server error within ~0.9s.

However, other models (like deepseek-v4-flash, qwen3.7-plus, qwen3.8-max) on the exact same endpoint work perfectly fine (HTTP 200 OK).

Request Details

• Endpoint: https://opencode.ai/zen/go/v1/chat/completions
• Model: gpt-5.6-luna
• Method: POST
• CF-RAY: a32ae0deaaf8312c-LAX
• Server: cloudflare

Response Payload

{
  "type": "error",
  "error": {
    "type": "error",
    "message": "Internal server error"
  }
}

### Plugins

None (Direct API call via /zen/go/v1)

### OpenCode version

Latest (API Gateway /zen/go/v1)

### Steps to reproduce

1. Send a standard HTTP POST request to `https://opencode.ai/zen/go/v1/chat/completions` with Authorization Bearer header.
2. Set request body payload to `{"model": "gpt-5.6-luna", "messages": [{"role": "user", "content": "ping"}]}`.
3. The gateway immediately responds with `HTTP 500 Internal server error` within ~0.9s.
4. If changing model to `deepseek-v4-flash` or `qwen3.7-plus` with the same headers, it returns `HTTP 200 OK` successfully.

### Screenshot and/or share link

<img width="797" height="417" alt="Image" src="https://github.com/user-attachments/assets/32a2e7a6-a801-4a3f-a460-a2266eb8aaa0" />

### Operating System

Windows 11

### Terminal

Windows Terminal / PowerShell