AIToday
Large Language ModelsAI Business & IndustryTHE DECODERPublished: Apr 26, 2026, 22:01 JST1 min read

OpenAI folds Codex coding model into GPT-5.5, discontinuing separate programming AI

OpenAI folds Codex coding model into GPT-5.5, discontinuing separate programming AI

3 Key Points

  1. OpenAI merged its dedicated Codex code-writing model into the main GPT-5.5 language model as of GPT-5.4 (released in early 2025). Version 5.3, shipped in early February, was the last standalone Codex release. This repeats a pattern: OpenAI shut down the original Codex in 2023, then revived it in May 2025 as Codex-1 before folding it back into the main model.

  2. GPT-5.5 requires fewer tokens (units of text the model processes) than GPT-5.4 to complete the same coding tasks, meaning faster responses and lower computing costs for the same work. However, despite this efficiency gain, OpenAI raised API pricing by approximately 20 percent, so users will pay more overall.

  3. Developers who rely on dedicated coding AI for programming tasks will now get those features built into GPT-5.5's general-purpose model instead of a separate tool. GPT-5.5 emphasizes agentic coding—where AI autonomously completes coding work with minimal human oversight—and improved ability to control computers, which may shift how teams structure code review and automation workflows.

Ask the AI about this article →

Get the latest Large Language Models news every morning

For example, today's edition would include:

  • Meta's Muse Spark 1.3 matches GPT-5.6-Sol, claims #3 modelLatent Space · 3h ago
  • CABiNet vs YOLO26-sem on UAVid: Accuracy, Compute, and GPU Latencyr/MachineLearning · 3h ago
  • C++ PCN library nears backprop accuracyr/MachineLearning · 3h ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleResearch reveals GPT, Gemini, and Claude all hallucinate when counting large datasets—but a structured protocol called KIS fixes the problem