AI NewsDev toolsAnnouncement

Codex 0.171.0 makes GPT-6.2 Sol the default and adds remote exec

Codex 0.171.0 adds GPT-6.2 Sol and Astra, with Sol as the default. The TUI workspace is on by default, and codex exec can now run on a remote server.

AI News

Editorial2 min read

LinkedInX
Codex release page on GitHub for version 0.171.0

Image: OpenAI

Why it mattersOpening a new chat now uses Sol unless you change it. The remote exec flag lets you start a long job on a server and pick it up later.

A coding agent you use every day gets rewritten by its defaults. Which model it opens with. Whether it runs on your laptop or a server. Whether a second pane stays on when you quit. OpenAI's Codex has a new release that changes all three. Version 0.171.0 is out on the project's GitHub.

The release notes list six new features and six bug fixes. Three of the features change what Codex does before you touch a setting.

GPT-6.2 Sol is now the default

OpenAI says it added GPT-6.2 Sol and Astra to the model catalogs. Sol is now the default. A new chat opens with Sol unless you pick another model. Chats you already have keep their picked model. The change does not rewrite an open session. OpenAI published GPT-6.2 Sol and Astra on 17 November. Sol is the lower-cost model. Astra is the slower, stronger one. The default change picks the one most users want for daily tasks.

The TUI workspace is on by default

OpenAI says the TUI workspace is now on by default. The side pane holds searchable document previews, image previews, and subagent transcripts. In 0.168.0 you had to turn this on through features.tui_workspace. Now a second pane is there from the first run. You can search the open documents. You can scroll back through what a subagent did. You can look at images without leaving the window.

Remote exec lets a job move between machines

OpenAI says codex exec --remote <ADDR> can now start or resume a noninteractive session on an existing app server. The /rc and /remote commands manage this from the TUI. The TUI shows a pairing code for the device you want to attach. You can start a long job on a workstation and close the laptop. You can pick up the same session from a phone or another machine. This is the first version where a Codex session is not tied to the process that started it.

Other changes worth a line each

OpenAI says a custom MCP status line can be chosen or written through /statusline. Six fixes land with the features. Codex now keeps a draft, a queued follow-up, or an interrupted message with its own chat. Switching tabs does not lose them. A permission profile stays the same when the working directory changes. A late completion event cannot cancel the approval for a newer turn. The scrolling bug that inserted stray characters is gone. Enterprise MCP sign-in works for identity providers that issue ID tokens without a refresh token.

If you run Codex on a shared machine, the remote exec flag changes how you set work up. Start the server once. Then attach from anywhere the server is reachable. The pairing code holds the authentication. For a team that uses Codex every day, the default model change is the one to watch. A chat opened on 20 November is billed at the Sol rate. Check your model setting before you expect the old one.

Source

openai/codex release rust-v0.171.0, GitHub.

SourceOpenAI

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

Share
LinkedInX
Start a project