server_error with OpenAI provider in OpenCode CLI, while direct Codex CLI works fine #23956

Question

Hi OpenCode team,

I want to ask whether this is a known compatibility/integration issue with OpenAI/Codex in OpenCode, especially in agent/streaming flow.

Problem

When I use my GPT/Codex account through OpenCode CLI, I often get this error:

{"type":"error","sequence_number":2,"error":{"type":"server_error","code":"server_error","message":"An error occurred while processing your request. You can retry your request, or contact us through our help center at help.openai.com if the error persists. Please include the request ID ac4ddf64-7ea0-49d7-8382-b8b6672310fc in your message.","param":null}}

What is strange is:

• This happens in OpenCode CLI
• But when I use Codex CLI directly, I do not get this error

So it seems like the issue may be related to the OpenCode/agent integration layer rather than Codex itself.

Environment

• OS: Windows
• OpenCode version: 1.14.20
• Provider: OpenAI
• Model: openai/gpt-5.4
• Terminal: hyper, windows terminal

Reproduction

1. Open OpenCode CLI
2. Use provider/model: gpt-5.4
3. Run an agent task or prompt
4. Wait for response streaming
5. Error appears very early with sequence_number: 2

Expected behavior

The request should complete normally, similar to when using Codex CLI directly.

Actual behavior

OpenCode intermittently returns server_error, while the same account/model workflow works in Codex CLI direct usage.

What I already tried

• Retried multiple times
• Confirmed the account/model works in Codex CLI directly
• Observed similar issue in another agent-style terminal environment
• Reconnected auth

Questions

• Is this a known issue with OpenAI provider support in OpenCode?
• Could this be related to streaming/event handling in agent mode?
• Is there any recommended config, provider setting, or model format I should change?
• Do you want me to provide debug logs for this?

Debug information

Please let me know which logs are most useful

Thanks.