OpenAI AI Updates: April 22, 2026

1. OpenAI Ships ChatGPT Images 2.0 With Reasoning and Web Search

OpenAI launched GPT Image 2, branded in product as ChatGPT Images 2.0, adding a thinking step that can call web search before rendering. With thinking enabled, one prompt produces up to eight consistent images, and the model handles small text, UI elements, and non-Latin scripts (Japanese, Korean, Hindi, Bengali) substantially better than its predecessor. Output goes up to 2K across aspect ratios from 3:1 to 1:3. Basic quality improvements reach all users; extended thinking outputs are gated to Plus, Pro, and Business. API access is gpt-image-2 with token-based pricing between $8 and $30 per million tokens depending on input and output type. Source

2. OpenAI Builds Out Codex Enterprise With Consulting Partnerships

OpenAI. Codex hit four million weekly active users and the company is formalizing how big companies deploy it, announcing Codex Labs and partnerships with major consulting firms to cover the deployment lifecycle from pilot through scaled rollout. The framing treats Codex less as a standalone product and more as a platform layer the consulting partners wrap with playbooks, governance, and customization for specific enterprise stacks. Source

3. Codex Chronicle Records Your Screen to Build Local Memory

OpenAI. Codex added Chronicle, an opt-in preview for ChatGPT Pro users on macOS that records screen activity in the background, summarizes it into Markdown memory files stored locally, and deletes the raw recordings after six hours. OpenAI’s own documentation flags the risks bluntly: rapid rate-limit consumption, a larger prompt-injection surface because anything visible on screen is now input, and unencrypted memory files on disk. The feature is not offered in the EU, UK, or Switzerland. Source