DeepSeek V4 Flash Official API Launches Public Beta: A Complete Guide to Codex Configuration
DeepSeek V4 Flash's official API is now in public beta, featuring native Responses API support and Codex adaptation. This guide covers setup steps for...
DeepSeek has officially released the public beta of its V4 Flash API, which includes native support for the Responses API and is optimized for Codex integration. This allows Codex users to easily leverage the DeepSeek V4 Flash model for their AI tasks.
Before configuring, users need three prerequisites: a Codex CLI or client (verify via codex --version), an existing ~/.codex directory (ensured by running Codex at least once), and a DeepSeek API Key (created on the DeepSeek Platform).
Windows users can set up with a PowerShell command: irm https://cdn.deepseek.com/api-docs/codex-deepseek-setup-en.ps1 | iex. Select option 1 in the script menu and enter your API Key. The script backs up the original config.toml, updates models.json to set deepseek-v4-flash as the default model, and points the provider to DeepSeek's Responses API. A test on Windows with Codex CLI 0.137.0 confirmed the script works.
For macOS/Linux users, use the bash command: bash <(curl -fsSL https://cdn.deepseek.com/api-docs/codex-deepseek-setup.sh). Configuration files are shared across Codex clients (CLI, desktop app, VS Code plugin), so restart clients after setup to clear old cache.
To verify the configuration, run codex doctor --json and check four metrics: config.load (OK), model (deepseek-v4-flash), provider (deepseek), and wire API (responses). A local test passed these checks, but note that an invalid demo key was used—this confirms configuration recognition, not real request performance or costs.
Key management tips: Avoid committing config.toml or screenshots to Git/public platforms. Team environments should use individual keys for easy revocation. To roll back, re-run the script and select option 3 (restores from ~/.codex/backup-deepseek; check backup timestamps if you modified configs manually).
After setup, review call costs. Codex Agent tasks have long contexts and tool calls, so compare DeepSeek V4's prices across channels and test token consumption with the same task to make informed choices.
Sources