Skip to content
AImpact
IT EN
Landmark Foundation Models · 1 min read

GPT-5.5: OpenAI shifts ChatGPT toward an "agent runtime" paradigm

In one sentence OpenAI releases GPT-5.5, GPT-5.5 Thinking, and GPT-5.5 Pro: designed as an "agent runtime" for persistent multi-step workflows. 23% more factually correct vs GPT-5.4. File Library, side-by-side shopping, improved image gen.

Verified Official source
ShareLinkedInX
Reading level

On April 23, 2026 OpenAI releases GPT-5.5 (plus Thinking and Pro variants). The strategic direction is right in the launch headline: "smartest and most intuitive to use model yet, and the next step toward a new way of getting work done on a computer".

Translation: OpenAI stops thinking about ChatGPT as a "chatbot" and starts thinking about it as an agent runtime — an environment where you delegate long tasks and the AI carries them through over days, keeping state, invoking tools, taking initiative.

Three new products at the same time:

  • File Library: ChatGPT remembers files you showed it across different conversations;
  • Side-by-side shopping: product comparison built in (a move toward "super-app");
  • Improved image gen: a rival to Nano Banana 2.

Concrete numbers: 23% more likely to be factually correct vs GPT-5.4, factual errors down 3%. Available to Plus/Pro/Business/Enterprise (not free tier). API ships the day after (April 24) with extra safeguards.

Companies

OpenAI

Tools

GPT-5.5, GPT-5.5 Pro, GPT-5.5 Thinking, ChatGPT

Tags

OpenAIGPT-5.5ChatGPTAgentsCodex

Sources