Skip to content
CC Switch WikiUnofficial Community Wiki

This model is not supported when using X-OpenAI-Internal-Codex-Responses-Lite - Codex

Quick fix

Comment out model_provider = custom in config.yaml to go back to the official OpenAI provider.

Symptom

Section titled “Symptom”
Error output
This model is not supported when using X-OpenAI-Internal-Codex-Responses-Lite.

Cause

Section titled “Cause”

The error is triggered by the cc-switch "unified history" feature. With it on, cc-switch sets `model_provider` to `custom` to take over session management. But Codex changed or tightened its internal model identifiers (`X-OpenAI-Internal-Codex-Responses-Lite`, for instance) in an update, so on the custom provider path Codex refuses to support the model currently configured.

This is not data loss but a configuration conflict. Some users report that after resolving the error, history may come back incompletely or a long conversation may disappear after a restart — usually because cc-switch modified the Codex internal index or session metadata, leaving the multi-layer indexes of Codex Desktop (SQLite threads, session_index.jsonl, and so on) out of alignment.

Fix

Section titled “Fix”

  1. Open the cc-switch config file, find the `model_provider = custom` line, and comment it out so Codex falls back to the official OpenAI provider.

Affected Versions

Section titled “Affected Versions”
ToolCodex
Version3.16.4
PlatformsWindowsmacOS

Source Issues

Section titled “Source Issues”

This page is distilled from 2 real issues

FAQ

Section titled “FAQ”
Why did only part of my conversations come back after turning unified history off?
Because cc-switch modified the Codex user configuration and index structure. Switching back to the official provider can leave the multi-layer Codex indexes (the active SQLite, session_index.jsonl) not fully in sync, so some sessions are filtered out or fall outside the sort window. Back up the ~/.codex directory first, then try repairing the index.
A long conversation disappeared after restarting Codex. Is the data still there?
The data is usually not deleted, just invisible in the UI. This is a "history visibility" problem, rooted in session metadata not matching the index logic of the newer Codex version. Do not edit the JSONL or SQLite files by hand — back up ~/.codex and ~/.cc-switch completely first, then follow the relevant investigation flow to restore index consistency.
Besides commenting out the custom provider, is there another temporary workaround?
You can switch to an officially supported stable model version (gpt-5.4, for instance), or go back to the genuine built-in openai provider without the cc-switch routing layer. But the most fundamental answer is still to avoid third-party plugin configurations that modify the internal Codex provider state while Codex is being updated.
Section titled “Related problems”
Codex provider API keys overwrite each other after a restart or switchEnable CC Switch local proxy takeover, or use experimental_bearer_token in config.toml, to bypass the auth.json sync defect.7 issues mergedCodex conversation history disappears after switching providerAfter upgrading to v3.16.0 or above, CC Switch automatically migrates third-party provider history into the custom bucket; original files are backed up.2 issues mergedapi failed deserialize json body intoDepending on the error: roll Claude Code back, upgrade cc-switch, turn the local proxy off, or stop sending unsupported fields.7 issues merged

This is an unofficial community wiki with no affiliation to the cc-switch authors or the project itself. Its content is compiled from the project's public GitHub issues. This site distributes no software.