Skip to content
CC Switch WikiUnofficial Community Wiki

Codex model catalog template `gpt-5.5` not found CC Switch

Quick fix

Regenerate ~/.codex/models_cache.json and keep the Codex ChatGPT login, so CC Switch can write the route and show DeepSeek models.

Symptom

Section titled “Symptom”
Error output
热切换失败: 写入 Codex 配置失败: Codex model catalog template `gpt-5.5` not found. Please start Codex once so models_cache.json is available, or ensure the `codex` CLI is on PATH.
English translation

Hot switch failed: failed to write Codex config: Codex model catalog template `gpt-5.5` not found. Please start Codex once so models_cache.json is available, or ensure the `codex` CLI is on PATH.

Cause

Section titled “Cause”

Codex's model catalogue `models_cache.json` is the template source CC Switch uses when writing Codex routing config. On a fresh Codex install with no login, or with the `codex` CLI missing from PATH, `~/.codex/models_cache.json` is never generated, so CC Switch fails to find the `gpt-5.5` template when enabling local routing or switching provider. Officially this was declared fixed in v3.16.0/v3.16.1, but several users still reproduce it after upgrading, so generating the cache by hand remains the method that works in practice.

A second presentation — the model list showing only 自定义 (custom) or GPT models with no `deepseek-v4-flash`/`deepseek-v4-pro` — comes from Codex desktop hiding third-party models and falling back to a default model when the official ChatGPT login is not preserved. That has nothing to do with the local cache. Some problems also come from CC Switch cloning the GPT template's vision and `apply_patch` capability fields when generating `~/.codex/cc-switch-model-catalog.json`, breaking AppShots and apply_patch. Those share a root and need editable templates (#6451) to fix properly.

Fix

Section titled “Fix”

  1. Run `codex debug models --bundled > ~/.codex/models_cache.json` in a terminal to generate a cache file containing `gpt-5.5`, `gpt-5.4`, and the other models.

  2. On a fresh Codex install, start it normally and sign in to a ChatGPT account once so `models_cache.json` is generated under `~/.codex`. If it still is not, drop a community-shared `models_cache.json` into `~/.codex/` by hand.

  3. Keep the official ChatGPT login in Codex desktop, then go back to CC Switch and enable local routing or switch endpoint — only then does the model picker show DeepSeek flash/pro.

  4. If it still shows only the old models, delete the old DeepSeek provider and recreate it from the built-in CC Switch DeepSeek preset (the preset is `deepseek-v4` from v3.16 on).

Affected Versions

Section titled “Affected Versions”
ToolCodex
Version3.16.0 - 3.16.5（cc-switch 受影响范围；官方称 3.16.0/3.16.1 已修复）
PlatformsWindowsmacOS

Source Issues

Section titled “Source Issues”

This page is distilled from 10 real issues

FAQ

Section titled “FAQ”
Why are the models still not found after running `codex debug models --bundled`?
First confirm `~/.codex/models_cache.json` actually contains `gpt-5.5`, `gpt-5.4`, and the rest. If Codex is not signed in or the `codex` CLI is off PATH, the file may be empty — sign in and make sure the CLI is available.
Why does Codex show only 自定义 (custom) with no deepseek-v4-flash/pro?
That is a Codex desktop limitation: without a preserved official ChatGPT login, the model picker hides third-party models and falls back to a default. Sign in to the official account once, then switch the CC Switch endpoint.
Why do AppShots or apply_patch misbehave?
The generated `~/.codex/cc-switch-model-catalog.json` inherits vision and `apply_patch` capability fields from the GPT template, and hand edits are overwritten on a switch. #6451 tracks editable templates; for now the only option is a manual temporary correction.
Section titled “Related problems”
api 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 mergedRequest body format invalid error - Claude CodeUpgrade cc-switch and enable the Rectifier, or add MAX_THINKING_TOKENS=0, for compatibility with third-party API formats.4 issues mergedConsole Go: Upstream request failedThis 400 comes from the OpenCode Go gateway rejecting an incompatible payload. Restart the proxy to restore session mapping, or avoid WebSearch and images.3 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.