UPGPTs

English

中文

English

中文

Appearance

Connect Codex to DeepSeek-V4-Flash (2026)｜One-Click CLI Script + CC Switch, Step by Step ​

DeepSeek-V4-Flash is fast and its agent capabilities are noticeably stronger than the previous generation. The good news: DeepSeek officially provides a one-click script to connect it to Codex — no code changes, no hand-written config files. Run one command, paste your API Key, done.

And you only configure it once: Codex CLI, the desktop app, and IDE extensions all pick it up.

TL;DR: Connecting Codex to DeepSeek-V4-Flash takes two steps — ① create an API Key on the DeepSeek platform; ② run DeepSeek's official one-click script (or use the GUI tool CC Switch). About 3 minutes, zero code. Prefer the command line? See Method 1. Prefer switching in the desktop app? See Method 2.

Contents:

  1. What Is DeepSeek-V4-Flash?
  2. What to Prepare Before You Start
  3. Method 1. Official One-Click Script
  4. Method 2. CC Switch GUI
  5. Verify the Configuration
  6. FAQ
  7. Related Reading

What Is DeepSeek-V4-Flash? ​

DeepSeek-V4-Flash is DeepSeek's fast-tier model, built for speed. Compared with V4-Pro Preview, its agent capabilities are noticeably stronger, which makes it a good fit for autonomous coding agents like Codex — reading code, editing files, and running tests.

Once connected, Codex's underlying model becomes DeepSeek-V4-Flash: you keep the familiar Codex interface, but the heavy lifting is done by DeepSeek.

What to Prepare Before You Start ​

  • ☐ Codex CLI 0.144.0 or newer: older versions won't recognize DeepSeek's official model catalog
  • ☐ A DeepSeek account and API Key: create the key on the DeepSeek platform
  • ☐ A terminal with internet access: Terminal on macOS/Linux, PowerShell on Windows

First, Check the Codex CLI Version ​

Run this in your terminal:

bash
codex --version

Run codex --version in the terminal to check your Codex CLI version

DeepSeek's official model catalog requires Codex CLI 0.144.0 or newer. If yours is older, update first:

bash
codex update

Method 1. Official One-Click Script ​

The official one-click script is the recommended way — just three steps: run the script → pick 1 → paste your API Key.

macOS and Linux ​

Run in your terminal:

bash
bash <(curl -fsSL https://cdn.deepseek.com/api-docs/codex-deepseek-setup.sh)

Windows ​

Run in PowerShell:

powershell
irm https://cdn.deepseek.com/api-docs/codex-deepseek-setup-en.ps1 | iex

The demo below uses Windows. After the script runs, DeepSeek asks what you want to do:

Menu shown after running the DeepSeek one-click setup script

To connect the latest DeepSeek-V4-Flash, pick 1:

Enter 1 in the script menu to connect DeepSeek-V4-Flash

The script then asks for your API Key — paste the one you created on the DeepSeek platform and press Enter:

Paste the API Key you created on the DeepSeek platform

That's it — the connection is done:

Confirmation that the connection was successful

Method 2. CC Switch GUI ​

Don't want to touch the command line? Use CC Switch, a GUI tool that switches model providers right inside the Codex desktop app (ChatGPT desktop).

  1. Open CC Switch and select the Codex tab

CC Switch main window

Select the Codex tab in CC Switch

  1. Click + to add a provider, choose the DeepSeek preset, fill in your API Key, and pick DeepSeek-V4-Flash

Add a provider in CC Switch

Choose the DeepSeek preset, enter the API Key, and select DeepSeek-V4-Flash

  1. In Advanced Options, the upstream format defaults to routing, which is fine. If you'd rather not use routing, pick Native — but you'll need CC Switch newer than v3.19.1 to see that advanced option.

Upstream format setting in CC Switch advanced options

  1. Save, then fully quit and restart Codex (ChatGPT desktop). If you see DeepSeek in the bottom-left corner after restarting, the config is working.

Verify the Configuration ​

In Codex CLI, type:

/model

If DeepSeek V4 Flash appears in the model list, the connection is live.

FAQ ​

Q: What version of Codex do I need to connect DeepSeek-V4-Flash? A: Codex CLI 0.144.0 or newer. If yours is older, run codex update first — otherwise DeepSeek's official model catalog won't be recognized.

Q: What's the official one-click setup command? A: macOS/Linux: bash <(curl -fsSL https://cdn.deepseek.com/api-docs/codex-deepseek-setup.sh). Windows PowerShell: irm https://cdn.deepseek.com/api-docs/codex-deepseek-setup-en.ps1 | iex.

Q: After one setup, which Codex clients can use it? A: Codex CLI, the desktop app, and IDE extensions. One setup is shared by every Codex client form.

Q: How do I confirm DeepSeek-V4-Flash is connected? A: Type /model in Codex CLI — if DeepSeek V4 Flash appears in the list, it's connected. With CC Switch, fully restart Codex and check for "DeepSeek" in the bottom-left corner.

Q: CC Switch upstream format: routed or native? A: The default routed mode is fine. Choose Native only if you don't want routing, and you need CC Switch newer than v3.19.1 for that advanced option.