Skip to content
NewsProduct launch

GitHub Copilot CLI v1.0.81-8 adds Grok 4.6 xhigh reasoning effort and fixes plugin loading

· by Pondero Newsdesk

The short version

GitHub shipped Copilot CLI v1.0.81-8 on August 23, 2026, adding xhigh reasoning effort for Grok 4.6 and fixing local plugin loading so directory edits take effect on restart without a reinstall.

GitHub Copilot CLI v1.0.81-8 adds Grok 4.6 xhigh reasoning effort and fixes plugin loading

Copilot CLI picked up a Grok 4.6 upgrade on August 23. Version 1.0.81-8 adds xhigh as a selectable reasoning effort level, fixes a plugin reload behavior that forced developers to reinstall locally edited extensions rather than simply restarting the session, and patches a stale-settings bug that affected enterprise sign-outs.

What changed

The release notes for v1.0.81-8 list three shipping changes, per the Copilot CLI releases page.

First, xhigh reasoning effort is now available when using Grok 4.6. Reasoning effort controls how long the model deliberates before responding; xhigh is the highest tier in the CLI's configuration. It is suited for multi-step coding problems and longer debugging sessions where depth matters more than response speed.

Second, the plugin loading behavior changed. Path-sourced plugins in a local directory-based marketplace previously required a /plugin update command after every edit. They now load live from their actual directory and reflect changes on /restart or at the start of a new session. Skills and custom agents defined in registered directories are also discovered automatically under the same logic.

Third, a sign-out fix landed. Signing out of an account now clears its cached enterprise managed settings. Previously, signing back in could re-apply a stale policy from before the sign-out instead of fetching the current one.

Context

Grok 4.6 became available across eight Copilot surfaces on August 14, per the GitHub changelog. Those surfaces are VS Code, Visual Studio, Copilot CLI, the Copilot cloud agent, the Copilot app, JetBrains, Xcode, and Eclipse. The August 14 rollout covered the model's basic selectable operation across all of them; the August 23 CLI release adds the xhigh effort tier as a distinct configuration option.

GitHub described the model as showing strong results in internal testing on terminal-based coding tasks requiring sustained reasoning and tool use. Grok 4.6 is billed at provider list pricing under Copilot's usage-based billing. Copilot Business and Enterprise administrators must turn on the Grok 4.6 policy in settings before it appears for their users; the policy ships off by default.

Why it matters

The xhigh reasoning effort setting fills a gap in how developers could direct Grok 4.6's capabilities at the terminal. Since the August 14 rollout, the model was available in the CLI but without the effort configuration that makes it practical for deep tasks. Setting it to xhigh is the right choice when the problem involves planning across many files, tracing bugs through multiple services, or scaffolding a feature that requires the model to hold a lot of context before writing a single line.

The plugin reload fix is the more immediately practical change for anyone building Copilot extensions locally. Tight edit-test cycles at the terminal now work as expected: edit the plugin, restart, see the change.

Enterprise teams using GitHub Copilot who have seen unexpected policy behavior after re-authentication should apply v1.0.81-8. The sign-out fix addresses the specific case where a cached policy from a previous session survived a sign-out and re-login.

What to watch

A weekly GitHub Copilot changelog entry covering these CLI changes had not appeared in the August 2026 index as of August 24. Those entries typically follow a release by a few days. Whether xhigh reasoning effort for Grok 4.6 extends to IDE surfaces such as VS Code has not been announced.

Sources