A pelican for GPT-5.5 via the semi-official Codex backdoor API

2026-04-23 · Source: Simon Willison's Weblog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, short

Summary

OpenAI released GPT-5.5 on April 23, 2026, making it available to paid ChatGPT subscribers and via OpenAI Codex, but not yet through its official API. This new model is described as fast, effective, and highly capable, though its specific API pricing will be $5 per 1M input tokens and $30 per 1M output tokens, double that of GPT-5.4. A new plugin, `llm-openai-via-codex`, allows users to access GPT-5.5 through their existing Codex subscriptions, leveraging a "semi-official" backdoor API endpoint that OpenAI has publicly endorsed for integration with third-party tools like OpenClaw and Pi. Benchmarking with a "pelican" SVG generation task showed improved results with higher reasoning effort, increasing from 39 to 9,322 reasoning tokens.

Key takeaway

For AI Engineers and developers seeking early access to GPT-5.5, you can integrate with the model immediately using the `llm-openai-via-codex` plugin and your existing OpenAI Codex subscription. This bypasses the official API delay and allows you to experiment with GPT-5.5's capabilities, including its improved performance with higher reasoning effort, before general API availability. Be mindful of the increased pricing for GPT-5.5 once the official API is released.

Key insights

OpenAI's GPT-5.5 is accessible via a semi-official Codex API, enabling early integration despite no public API release.

Principles

• OpenAI supports third-party integration via Codex API.
• Increased reasoning effort improves generative model output quality.

Method

Access GPT-5.5 through a Codex subscription by installing the `llm-openai-via-codex` plugin for LLM, then use `llm -m openai-codex/gpt-5.5` for prompting.

In practice

• Use `llm install llm-openai-via-codex` to enable access.
• Apply `-o reasoning_effort xhigh` for complex generations.

Topics

• GPT-5.5
• OpenAI Codex API
• LLM Plugin
• Agent Harnesses
• Model Pricing

Code references

• openai/codex
• simonw/llm-openai-via-codex

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• OpenAI kills its dedicated coding model Codex again, folding it into GPT-5.5 — OpenAI has integrated its dedicated Codex coding model into the main GPT-5.x series, enhancing agentic coding capabilities.
• Sign of the future: GPT-5.5 — AI capabilities are rapidly accelerating across models, applications, and tool integration.
• llm-openai-via-codex 0.1a0 — A new tool allows OpenAI model access using existing Codex CLI credentials.
• llm 0.32a2 — OpenAI's GPT-5 class models now use `/v1/responses` for interleaved reasoning, enhancing tool call visibility.
• Introducing OpenAI’s GPT-5.4 mini and GPT-5.4 nano for low-latency AI — OpenAI has introduced GPT-5.4 mini and GPT-5.4 nano, smaller variants of GPT-5.4 optimized for developer workloads prioritizing low latency, cost savings, and agentic design.
• Codex + GPT-5.5 = SUPER APP! Build and Do ANYTHING! — OpenAI's GPT 5.5 and Codex offer a superior, more efficient AI coding and automation workflow.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.