求助!Codex app 没法显示除了 gpt 以外的任何模型

不可回收垃圾 2026-07-09 15:38 1

如题,今天看着 grok 发布了,搞了一个账号在 cpa 上反代,然后想在 codex 里使用,然后使用 ccs 配置, 并且也有model_catalog_json ,在 codex cli 上,使用/model 可以正常显示,但是在 codex app 上,却只能显示和选择gpt 的模型,如图:




然后下面是config.toml的配置,


model_provider = "custom"
model = "grok-4.5"
model_reasoning_effort = "high"
disable_response_storage = true

notify = ["/Users/mac/.codex/computer-use/Codex Computer Use.app/Contents/SharedSupport/SkyComputerUseClient.app/Contents/MacOS/SkyComputerUseClient", "turn-ended"]
model_catalog_json = "cc-switch-model-catalog.json"

[model_providers.custom]
name = "custom"
base_url = "https://xxxxx"
wire_api = "responses"
requires_openai_auth = true

然后是 model_catalog_json的配置:


{
"models": [
{
"additional_speed_tiers": [],
"availability_nux": null,
"base_instructions": "You are Codex, a coding agent. You and the user share the same workspace and collaborate to achieve the user's goals.",
"context_window": 128000,
"default_reasoning_level": "high",
"default_reasoning_summary": "none",
"description": "gpt-5.5",
"display_name": "gpt-5.5",
"effective_context_window_percent": 95,
"experimental_supported_tools": [],
"input_modalities": [
"text"
],
"max_context_window": 128000,
"priority": 1000,
"service_tiers": [],
"shell_type": "shell_command",
"slug": "gpt-5.5",
"support_verbosity": false,
"supported_in_api": true,
"supported_reasoning_levels": [
{
"description": "Disable Thinking",
"effort": "none"
},
{
"description": "Enabled Thinking",
"effort": "high"
}
],
"supports_image_detail_original": false,
"supports_parallel_tool_calls": false,
"supports_reasoning_summaries": true,
"supports_search_tool": false,
"truncation_policy": {
"limit": 10000,
"mode": "bytes"
},
"upgrade": null,
"visibility": "list"
},
{
"additional_speed_tiers": [],
"availability_nux": null,
"base_instructions": "You are Codex, a coding agent. You and the user share the same workspace and collaborate to achieve the user's goals.",
"context_window": 128000,
"default_reasoning_level": "high",
"default_reasoning_summary": "none",
"description": "grok-4.5",
"display_name": "grok-4.5",
"effective_context_window_percent": 95,
"experimental_supported_tools": [],
"input_modalities": [
"text"
],
"max_context_window": 128000,
"priority": 1001,
"service_tiers": [],
"shell_type": "shell_command",
"slug": "grok-4.5",
"support_verbosity": false,
"supported_in_api": true,
"supported_reasoning_levels": [
{
"description": "Disable Thinking",
"effort": "none"
},
{
"description": "Enabled Thinking",
"effort": "high"
}
],
"supports_image_detail_original": false,
"supports_parallel_tool_calls": false,
"supports_reasoning_summaries": true,
"supports_search_tool": false,
"truncation_policy": {
"limit": 10000,
"mode": "bytes"
},
"upgrade": null,
"visibility": "list"
},
{
"additional_speed_tiers": [],
"availability_nux": null,
"base_instructions": "You are Codex, a coding agent. You and the user share the same workspace and collaborate to achieve the user's goals.",
"context_window": 128000,
"default_reasoning_level": "high",
"default_reasoning_summary": "none",
"description": "gpt-5.4",
"display_name": "gpt-5.4",
"effective_context_window_percent": 95,
"experimental_supported_tools": [],
"input_modalities": [
"text"
],
"max_context_window": 128000,
"priority": 1002,
"service_tiers": [],
"shell_type": "shell_command",
"slug": "gpt-5.4",
"support_verbosity": false,
"supported_in_api": true,
"supported_reasoning_levels": [
{
"description": "Disable Thinking",
"effort": "none"
},
{
"description": "Enabled Thinking",
"effort": "high"
}
],
"supports_image_detail_original": false,
"supports_parallel_tool_calls": false,
"supports_reasoning_summaries": true,
"supports_search_tool": false,
"truncation_policy": {
"limit": 10000,
"mode": "bytes"
},
"upgrade": null,
"visibility": "list"
},
{
"additional_speed_tiers": [],
"availability_nux": null,
"base_instructions": "You are Codex, a coding agent. You and the user share the same workspace and collaborate to achieve the user's goals.",
"context_window": 128000,
"default_reasoning_level": "high",
"default_reasoning_summary": "none",
"description": "gpt-5.4-mini",
"display_name": "gpt-5.4-mini",
"effective_context_window_percent": 95,
"experimental_supported_tools": [],
"input_modalities": [
"text"
],
"max_context_window": 128000,
"priority": 1003,
"service_tiers": [],
"shell_type": "shell_command",
"slug": "gpt-5.4-mini",
"support_verbosity": false,
"supported_in_api": true,
"supported_reasoning_levels": [
{
"description": "Disable Thinking",
"effort": "none"
},
{
"description": "Enabled Thinking",
"effort": "high"
}
],
"supports_image_detail_original": false,
"supports_parallel_tool_calls": false,
"supports_reasoning_summaries": true,
"supports_search_tool": false,
"truncation_policy": {
"limit": 10000,
"mode": "bytes"
},
"upgrade": null,
"visibility": "list"
}
]
}

目的就是想在 codex app 上使用 grok 模型,在 opencode 之类的模型倒是可以使用,但是使用 codex 习惯了,有路过的佬可以帮忙看看吗? ^-^

最新回复 (16)
  • zzgg 07-09 15:39
    1

    切换一个新会话,应该就显示自定义了,就可以使用了

  • wangzhen3691 07-09 15:45
    2

    如果我没记错的话新版本的codex桌面版这几个候选模型是写死的,老版本我记得deepseek是可以显示的。


    不行就让它自己patch一下,加个选项的事

  • 不可回收垃圾 楼主 07-09 15:45
    3

    佬,还是不行,我重启了,然后也开了好几个新对话,一直都只会显示 gpt 的模型 ^-^

  • cuiyuyuan 07-09 15:46
    4

    要再cc里面启用路由,才可以加载出对应的模型

  • 不可回收垃圾 楼主 07-09 15:47
    5

    不行就让它自己patch一下,加个选项的事



    佬,这个意思是让 codex app 给自己搞一个选项上去吗?还是回退旧版本呢?

  • 不可回收垃圾 楼主 07-09 15:48
    6

    ccs 里我启用路由也尝试过,还是无法显示,感谢像是写死了 ^-^

  • wangzhen3691 07-09 15:48
    7

    让cli把codex桌面版的前端解包开,自己搞一个选项上去。



    想要什么自己加!凹凸曼不给造自己造!

  • leileio 07-09 15:49
    8

    我问过codex,它说可以自己加,但很麻烦,它劝我直接用cli,佬你问问codex,看它怎么说

  • cuiyuyuan 07-09 15:50
    9

    我试过加载的mimo模型,

  • YUhome 07-09 15:53
    10

    这个咋弄的啊?大佬分享下。现在没 gpt 啥都只能用了

  • 不可回收垃圾 楼主 07-09 15:55
    11

    有道理,但是这个看起来很费 token 的样子 ^-^,我试着捣鼓捣鼓,谢谢佬

  • 不可回收垃圾 楼主 07-09 15:57
    12

    这个我在 codex cli 是可以的,cli 可以选择其他模型,但是 codex app 里不知道为什么死活不行 ^-^

  • 孤寂地走 07-09 16:15
    13

    新版本需要保持登录态才能够显示自定义模型。

    如果是使用cc-switch,可以打开codex应用增加,切换第三方保留官方登录

  • cuiyuyuan 07-09 16:27
    14

    cc switch 里面配置好之后,有codex++(GitHub - BigPizzaV3/CodexPlusPlus: An enhanced tool for CodexApp, striving to make Codex better to use and more comfortable 一个CodexApp的增强工具,努力让Codex变得更好用更舒服 · GitHub )codex++供应商配置里面现在第三方导入cc里面配置,然后退出codex,在codex++里面重启,就能看到其他模型了

  • 不可回收垃圾 楼主 07-09 16:33
    15

    我去,这个可以啊,完美解决,谢谢佬 ^-^

  • MOYUN 07-11 18:23
    16

    俺也解决了,找了半天这在找到了方案哈哈哈 ^-^

* 帖子来源Linux.do
返回