In Codex CLI version 0.40 , OpenAI switched to gpt-5-codex as the default model and introduced several developer-focused improvements: autocompaction is now triggered when the session context reaches 220k tokens ; usage limits/quotas are now visible on the /status page; and a new set of /review commands are available for reviewing specific commits, comparing against a baseline branch, or passing custom review instructions. Additionally, the default timeout for MCP tool calls is set to 60s , which can be adjusted for each MCP service in config.toml.
This update aligns with recent official upgrades to the Codex ecosystem, including improved CLI and IDE/web collaboration, automated code review, and enhanced workflow transparency. Developers can upgrade to 0.40 via npm or Homebrew. Details of the changes and migration instructions can be found in the repository's release notes and developer center.
Frequently Asked Questions
Q: What is the default model for 0.40?
A: gpt-5-codex, it can be used without manual switching.
Q: When is automatic compression triggered?
A: Automatically when the context is close to 220k tokens to control the session size.
Q: What new capabilities does /review add?
A: It supports single commit review, baseline branch-based difference review, and custom review instruction mode.
Q: How do I view quotas and rate limits?
A: Check the current usage limit and consumption in /status of the CLI.
Q: How to adjust the timeout of MCP calls?
A: The default is 60 seconds ; this can be set individually for each MCP server in config.toml.tool_timeout_sec
