OpenAI is actively trying to seize the initiative from Claude Code, investing heavily in this effort.

Codex remains free for another month
https://openai.com/codex/
An extension of the original limited-time promo from February 2, 2026. Following the release of the Windows version of the Codex app, the promotion has been extended by another month; free ChatGPT accounts can now generate code until April 2. Plus accounts receive double limits.

Codex app for Windows and GPT‑5.4
https://openai.com/index/introducing-gpt-5-4/
OpenAI has finally introduced the Windows version of the Codex app and GPT‑5.4 - a new model that combines the coding capabilities of GPT-5.3-Codex with powerful reasoning. As usual, the model is more token-efficient, faster in iterations, and more proactive.

https://www.youtube.com/watch?v=8hNcRChDrNk

A specialized WinUI App skill has been added for Windows developers. You can now select different terminals and switch to WSL.

Starting from version 26.305, a fast mode has been introduced where GPT-5.4 operates 1.5 times faster while maintaining the same level of intelligence.

On the downside, the "Default open destination" list cannot be edited.

Reports suggest GPT-5.4 can view screenshots, control the mouse and keyboard, and run Playwright in Interactive mode for real-time visual debugging.

WebSocket Mode
https://developers.openai.com/api/docs/guides/websocket-mode/
This is a persistent connection for the Responses API, specifically designed for long agentic workflows with numerous tool calls (agentic coding, automation, orchestration). For coding agents, it significantly reduces iteration latency, offering up to 40% faster execution with 20+ tool calls.

The mode is built into the Codex App (macOS/Windows). In Codex-Spark, this mode is enabled by default. For other models, you need to add responses_websockets_v2 = true to the ~/.codex/config.toml configuration file (CLI version v0.110 will display an "Under-development features" warning).

#codex #chatgpt #newllmmodel