Vibeleaderboard
← All Intel
Intel / repo

codex rust-v0.171.0 — 0.171.0

Source
github.com
Author
github-actions[bot]Top Viber
Date

Context

Codex 0.171.0, a release of the open-source command-line version of OpenAI's coding , changes which model people get when they have not picked one: GPT-6.2 Sol replaces GPT-6.1 Sol as the default (release notes). The change, pull request #62106, adds GPT-6.2 Sol and GPT-6.2 Astra to the model list Codex ships with and to its Amazon Bedrock and Google Cloud Vertex AI lists, and makes Sol the default selection. Its startup test shows the effect: a fresh session with no model configured now displays GPT-6.2-Sol where it previously displayed GPT-6.1-Sol. The older model stays in the model picker, relabelled as a previous-generation model, and its in Codex's list rises to 500,000 tokens.

The evidence so far does not show that the new default codes better. OpenAI's API documentation lists GPT-6.2 Sol at $2 per million input tokens and $10 per million output tokens, the same price as GPT-6.1 Sol. On Artificial Analysis's coding agent index, which runs models inside Codex, GPT-6.2 Sol scored 3 points below GPT-6.1 Sol, a drop the firm traced to a question-answering test within the index, where the new model more often left key facts out of its answers. In isu-bench, where agents tune a slow web service with a fixed budget of load tests, GPT-6.2 Sol reached 14.1% of the estimated maximum and GPT-6.1 Sol 15.2%, from a single run each in different Codex versions, and the benchmark's author warns that small differences between single runs are not meaningful. Users who prefer the previous model can still select it.

The release also lets codex exec, the mode that runs Codex from scripts without a person at the keyboard, attach to an app server that is already running. According to pull request #61707, codex exec --remote with an address starts a new thread or resumes one by its ID on that server, using the server's configuration and credentials, with optional bearer-token authentication. It refuses configuration overrides the remote path does not support and reports an error if the server disconnects before the turn finishes. Elsewhere, the terminal interface's workspace, which shows searchable document previews, image previews and transcripts alongside the main conversation, is now on by default.

Key takeaways · AI-distilled
  • Codex 0.171.0 makes GPT-6.2 Sol the default selection in its model catalogs, with Astra added alongside it as an option, per OpenAI's release notes.
  • Switching or resuming sessions now keeps drafts, queued follow-ups and interrupted input attached to their original threads, and steering an active turn no longer drops earlier edits from the aggregate diff.
  • The full changelog hardens the Linux : seccomp restrictions now cover x32 syscalls, network sandbox filters block request_key, and an opt-in transparent TCP routing mode is available.
  • The release adds SPIFFE JWT-SVID support for workload identity authentication, and source builds and release tooling move to Rust 1.99.1.
Terms in this piece · Glossary
  • AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
  • context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
  • multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.
  • sandbox — An isolated environment where AI-generated code or agent actions run without being able to touch anything real.
More from github-actions[bot]
Recommended reads
Comments

Checking sign-in…

Loading comments…