GPT-5.4 Now Available in Codex
OpenAI has released GPT-5.4 as the recommended model for most Codex tasks, combining recent advances in reasoning, coding, and agentic workflows. This represents a significant capability upgrade for developers and teams using Codex for professional work.
Key Features
Native Computer-Use Capabilities: GPT-5.4 is the first general-purpose model with native computer-use capabilities built directly into Codex, enabling agents to interact with applications more naturally.
Extended Context Window: The model includes experimental support for a 1M context window, allowing it to handle substantially larger codebases and longer task sequences without context limitations.
Improved Tool Efficiency: GPT-5.4 delivers stronger tool use and tool search capabilities, helping agents more effectively discover and utilize the right tools for complex workflows.
Availability and Access
GPT-5.4 is available across the entire Codex ecosystem:
- Codex App: Select GPT-5.4 from the model selector in the composer
- CLI: Use
codex --model gpt-5.4to start a new thread or/modelduring a session (requires CLI 0.111.0 or later) - IDE Extension: Choose GPT-5.4 from the model selector in the composer
- Codex Cloud: Available on the web platform
- OpenAI API: Also accessible via direct API calls
Additional Updates
With the Codex CLI 0.111.0 release, Fast mode is now enabled by default with TUI header indicators showing the active mode (Fast or Standard). The js_repl can now dynamically import local .js and .mjs files, simplifying workspace script reuse.