Skip to documentation
deepthroat
DocsIntegrations
Terminal / editor agent

Codex

Connect Codex once with every text model in its model picker.

Responses API
On this pageBefore you startSetupConfigurationVerify & troubleshoot

Before you start

One deepthroat key connects Codex to the full text model catalog. Add credits or start an hourly plan on that key.

36 text models · one connectionComplete configuration included
All models
Get your API keyInstall Codex

Setup

  1. 1
    Download the model catalog below and save deepthroat-models.json in your ~/.codex folder (Windows: %USERPROFILE%\.codex).
  2. 2
    Merge the short TOML configuration into ~/.codex/config.toml. Keep top-level settings above provider tables.
  3. 3
    Set your API key with the terminal commands below, launch Codex, and open /model to choose from the full catalog.

Configuration

Starting model

All 36 models stay included. This only sets the initial selection.

Reasoning effort

Choose a supported level for this model. Higher effort can take longer.

deepthroat-models.json36 models included · no API key embedded
Download
~/.codex/config.toml
model = "gpt-6.1-sol"
model_provider = "deepthroat"
web_search = "disabled"
model_catalog_json = "deepthroat-models.json"

[model_providers.deepthroat]
name = "deepthroat"
base_url = "https://api.cheapai.gay/v1"
env_key = "deepthroatkey"
wire_api = "responses"
requires_openai_auth = false
supports_websockets = false

The catalog path in this TOML is relative to your user configuration folder. Keep both files in ~/.codex, then set your key and launch:

export deepthroatkey="YOUR_API_KEY"
codex

Use a recent Codex release with model_catalog_json support. HTTP Responses streaming and client tools are supported; image input is enabled only for models marked Vision on your selected node. Web search is opt-in and disabled in this starter configuration. See the Web search guide for API controls. WebSockets remain disabled.

Reasoning controls
Use /model to choose the model and its supported effort. The catalog includes Max and Ultra only for models that publish them. The effort selected above becomes your startup setting.
Switch models inside your app
Open /model to switch. Download the catalog again when new models are added.
Included models (36)
DeepSeek V4 Prodeepseek-v4-proModel default
DeepSeek V4.1 Flashdeepseek-v4.1-flashModel default
Uncensored DeepSeek V4.1 Flashuncensored-deepseek-v4.1-flashModel default
GPT-OSS 120Bgpt-oss-120bmedium
Muse Spark 1.3muse-spark-1.3Model default
Qwen3.7 Plusqwen3.7-plusModel default
Qwen3.8 Maxqwen3.8-maxModel default
Qwen3.7 Maxqwen3.7-maxModel default
Qwen3.6 Plusqwen3.6-plusModel default
Qwen3.5 Plusqwen3.5-plusModel default
Qwen3.8 Omni Flashqwen3.8-omni-flashModel default
Qwen3.5 Omni Plusqwen3.5-omni-plusModel default
Gemini 3.8 Flashgemini-3.8-flashlowmediumhigh
LongCat 2.0longcat-2.0Model default
LongCat 2.5 Previewlongcat-2.5-previewModel default
GPT-6 Astragpt-6-astralowmediumhighxhighmaxultra
GPT-6 Solgpt-6-sollowmediumhighxhighmax
GPT-6 Lunagpt-6-lunalowmediumhighxhighmax
GPT-5.6 Solgpt-5.6-sollowmediumhighxhighmax
GPT-5.6 Terragpt-5.6-terralowmediumhighxhighmax
GPT-5.6 Lunagpt-5.6-lunalowmediumhighxhighmax
Claude Sonnet 5claude-sonnet-5Model default
Claude Sonnet 4.6 (Thinking)claude-sonnet-4.6lowmediumhighxhighmax
Claude Haiku 4.5claude-haiku-4.5Model default
Claude Opus 4.5claude-opus-4.5lowmediumhighxhighmax
Claude Opus 4.6 (Thinking)claude-opus-4.6lowmediumhighxhighmax
Claude Opus 4.7claude-opus-4.7lowmediumhighxhighmax
Claude Opus 4.8claude-opus-4.8lowmediumhighxhighmax
Claude Opus 5claude-opus-5lowmediumhighxhighmax
Claude Opus 5.5claude-opus-5.5lowmediumhighxhighmax
Claude Sonnet 5.5claude-sonnet-5.5lowmediumhigh
GPT-6.1 Solgpt-6.1-sollowmediumhighxhighmaxultra
SWE-1.6swe-1.6Model default
Claude Haiku 5.5claude-haiku-5.5lowmediumhighxhighmax
DeepSeek V4.1 Flash 1017deepseek-v4.1-flash-1017lowmediumhigh
GLM 5.3 Flashglm-5.3-flashlowmediumhigh

Effort labels show levels available through this harness. These are the current published text models. Your key’s access rules and model availability still apply. Image-generation models use a separate API.

Verify & troubleshoot

Send a short prompt and check Activity. If models are missing, verify the list with the same API key:

bash
curl "https://api.cheapai.gay/v1/models" \
  -H "Authorization: Bearer YOUR_API_KEY"
  • 401: replace the placeholder with your actual key and check that it is enabled.
  • 402: top up that key or start an hourly plan in the wallet.
  • 404: refresh the model list and check the endpoint. Use https://api.cheapai.gay/v1 as the base URL.
  • Tool errors: select a tool-capable text model and read the tool calling guide.
Official Codex setup documentation
Build with your own key.Need help?