gpt 5.3 codex from OpenAI provider does not show its thinking

[lopi-py]

Description

Only the thinking block header is shown, but the actual content is not. This does not happen with gpt 5.2

Plugins

No response

OpenCode version

1.1.53

Steps to reproduce

1. Select gpt 5.3 codex from OpenAI
2. set variant to high or xhigh
3. ask questions

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• xhigh reasoning variant unavailable for GPT-5.3 Codex #12355: xhigh reasoning variant unavailable for GPT-5.3 Codex - also reports GPT-5.3 issues with reasoning/thinking variants
• GPT 5.2 thinking blocks missing #11273: GPT 5.2 thinking blocks missing - reports missing thinking blocks from OpenAI models on Windows

Feel free to ignore if none of these address your specific case.

[NamedIdentity]

Can confirm this issue.

I was hunting down the thinking blocks missing issue I reported a few weeks ago for ChatGPT via Plus oauth (#11273). Resolved that, but found Codex 5.3 remaining output these thinking blocks that are overly concise to the point of not really be a block of thinking.

[lopi-py]

I believe it will be fixed once gpt 5.3 codex is available via api

[MattesMrzik]

For me also the Github Copilot provided GPT-5-mini does not show thinking.

[diegonix]

Confirming the issue on OpenCode 1.2.20 with GPT-5.3-Codex models (high/xhigh reasoning effort).

Symptoms:

• Only thinking block headers display ("Thinking: ...")
• Content is truncated/hidden
• Works fine with GPT-5.2
• Configuration "include": ["reasoning.encrypted_content"] in opencode.json doesn't resolve it

Environment:

• OpenCode: 1.2.20
• Model: openai/gpt-5.3-codex
• OS: Linux (likely affects all platforms)

This appears to be a rendering issue in the TUI rather than a response parsing problem, since the header is visible but the reasoning content isn't displayed.

@rekram1-node - Would this be related to how streaming responses are buffered/rendered for 5.3's reasoning format vs 5.2? Happy to provide logs or help with debugging if needed.

Thanks for looking into this.

[josevelaz]

I believe im also seeing the same issue with 5.4

[iTzFaisal]

I'm seeing the same in GPT-5.3-Codex from GitHub Copilot in OpenCode. I can't see the thinking text at all !

I tried GPT-5.4 mini from GitHub Copilot in OpenCode and I can see the thinking text.

I've enabled /thinking or both.