Simon Willison Documents Codex API Access Route to GPT-5.5

Simon WillisonSimon Willison

· Updated

Simon Willison documented a method to access the new GPT-5.5 model through the Codex backend API using personal ChatGPT subscriptions. This discovery allows developers to bypass the delayed official API release and integrate the frontier model into local workflows immediately.

Simon Willison documented a route to access the gpt-5.5 model through the Codex backend API, bypassing the delayed official developer release. He describes using a personal ChatGPT subscription with a new llm-openai-via-codex plugin to call the model directly. This method enables the xhigh reasoning setting for complex tasks.
Plugin name
llm-openai-via-codex
Model ID
openai-codex/gpt-5.5
Reasoning mode
xhigh
Official API pricing (input)
$5 per 1M tokens
Official API pricing (output)
$30 per 1M tokens
Availability
Codex backend API via subscription

This finding highlights a demand for bridges between flat-rate subscriptions and API-driven workflows. Simon interprets this as a developer-friendly alternative to providers who block such access, mirroring the upgrade to the Codex platform. This follows a pattern of maintaining subscription-based access for individual developers.

To use this route, install the Codex CLI and the llm-openai-via-codex plugin to pick up existing authentication tokens. While this workaround is available now, the official API will eventually cost $5 per 1M input tokens—double the price of the GPT-5.4 release it succeeds.

View the full update on openai.com
Simon Willison
Simon Willison
@simonw
X

GPT-5.5 may not be in the official OpenAI API... but it's available via the apparently approved-of Codex API backdoor So I used that to make these pelicans (default and xhigh)! https://t.co/kIx2wnhpqD https://t.co/PP1JXhtEiM

3retweets101likes
View on X

Still wondering? A few quick answers below.

Simon Willison documented a method to access the new GPT-5.5 model through the Codex backend API using personal ChatGPT or Codex subscriptions. This discovery allows developers to bypass the delayed official API release and integrate the frontier model into local workflows immediately, even before the standard pay-per-token API is available.

The workaround uses the existing Codex backend API, which is the same mechanism used by OpenAI's official Codex CLI tool. By using authentication tokens stored by the Codex CLI install, developers can route prompts through their existing flat-rate subscriptions. This allows for model calls that include advanced features like the high-effort reasoning mode for complex tasks.

To use this method, you must first install the official Codex CLI and log in with an active OpenAI subscription. Next, install the LLM tool and Simon Willison's new plugin called llm-openai-via-codex. Once configured, you can run prompts against the gpt-5.5 model directly from your terminal using standard command-line instructions.

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

Share this update