GitHub Copilot changelog

Copilot Fusion adds local model routing in experimental preview

Copilot Fusion can now choose between supported on-device models and cloud models in Copilot CLI, the Copilot app and VS Code. MAI Code 1.1 Flash can be downloaded on supported Windows devices.

Who is affected:

Copilot users who want to run local models with automatic routing.

What to do:

Enable experimental local routing to try it.

Copilot Automations adds AI credit limits per run, day, week or month

You can set optional AI credit limits on each automation. Automations without limits keep running as before.

Who is affected:

Teams using Copilot Automations on Pro, Pro+, Max, Business and Enterprise plans.

What to do:

Set credit limits when creating or editing an automation.

GitHub Enterprise now includes Copilot for every licensed user, Standalone is $10

Enterprises working with a GitHub or Microsoft account team can get Copilot included with GitHub Enterprise. Users without an Enterprise license get Copilot Standalone at $10 per user per month, with no monthly AI credits per seat.

Who is affected:

Enterprise admins and users on GitHub Enterprise Cloud who manage Copilot seats and AI usage budgets.

What to do:

Talk to your sales team about moving an enterprise account to the new model, or keep your current plan.

Copilot Automations adds shared automations billed to the repository owner

Shared automations are visible to everyone with repository read access and billed to the repository owner instead of the creator. They are rolling out on Copilot Pro, Pro+, Max, Business, and Enterprise plans.

Who is affected:

Teams using Copilot Automations and organization admins who pay for AI credits.

What to do:

Org admins should enable the cloud agent billed to the organization policy to allow shared automations.

Copilot cloud sandboxes bill from November 2026 against a shared Actions entitlement

The included Actions compute entitlement becomes one shared, dollar-based pool instead of a fixed number of standard runner minutes. It covers standard runners, cloud sandboxes, and larger runners on Team and Enterprise Cloud. Billing for cloud sandbox usage begins in November 2026.

Who is affected:

Paid Copilot users and GitHub Free, Pro, Team and Enterprise Cloud accounts that use cloud sessions or Actions.

Action needed:

Review Actions budgets and usage, since higher-rate compute uses the entitlement faster.

GitHub Copilot Fusion becomes generally available with a 30% launch discount

Copilot Fusion routes and orchestrates models per task with Standard, High and Max effort levels. Users are billed per model used at standard rates, with a limited-time 30% discount on paid plans.

Who is affected:

All Copilot users in VS Code, the Copilot app, Copilot CLI and JetBrains, and admins who set model defaults.

What to do:

Update your client and select Copilot Fusion in the model picker.

Copilot Memory becomes generally available on paid plans

Copilot can remember personal preferences and repository conventions across sessions and clients. Copilot dreaming, which curates memories and proposes skill updates, launches in public preview and dream runs consume AI credits.

Who is affected:

Paid Copilot plan users and admins who want Copilot to carry context between sessions.

What to do:

Review memory settings and decide whether to opt in to dreaming.

Copilot code review adds Max effort level in public preview

Max joins Lite and Balanced as a third review effort level and runs several independent review agents across models from different providers. Copilot approvals are now generally available, off by default, and available on Pro, Pro+, Max, Business and Enterprise plans.

Who is affected:

Copilot code review users on paid plans, and admins who set review effort and approvals.

What to do:

Select Max for high-risk pull requests, and have admins turn on approvals where needed.

Copilot global default policy for features starts applying to unconfigured settings

GitHub began rolling out enforcement of the global default policy for Copilot Business and Enterprise through October 26. Generally available features and MCP server settings left unconfigured are set once to the chosen default of Enabled, Disabled or Let organizations decide. Explicit choices are kept.

Who is affected:

Copilot Business and Enterprise admins with unconfigured feature settings.

What to do:

Check AI Controls, Copilot, Default policy for new features before October 26.

Copilot removes Gemini 3.7 Flash, Grok 4.5 and older GPT models

On October 19, 2026 Copilot deprecated Gemini 3.7 Flash (use Gemini 3.8 Flash), Grok 4.5 (use Grok 4.7), and GPT-5.5, GPT-5.4, GPT-5.4 mini and GPT-5 mini.

Who is affected:

Copilot users and integrations pinned to those models.

Action needed:

Switch to the suggested alternatives. Admins may need to enable them.

Copilot to remove Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra on November 16

Copilot will deprecate Claude Opus 4.8 and Claude Opus 5 (replace with Claude Opus 5.5) and GPT-5.6 Terra (replace with GPT-6.1 Sol or GPT-6 Luna) on November 16, 2026. Opus 4.8 stays for individual annual-plan subscribers.

Takes effect Nov 16, 2026 · in 15 days

Who is affected:

Copilot users and integrations pinned to those models.

Action needed:

Move to the suggested alternatives before November 16, 2026.

Copilot adds HydraFusion Max multi-model option in research preview

HydraFusion, which combines several models in one Copilot response, now draws on Claude Opus 5.5, GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. The new HydraFusion Max uses higher reasoning effort for harder tasks. Both are research previews for Pro, Pro+, Business and Enterprise with preview features on.

Gemini 4 Argon available in GitHub Copilot for Pro+ and above

Gemini 4 Argon is GA in Copilot for Pro+, Max, Business and Enterprise users (not Pro), billed at provider list pricing.

Claude Haiku 5.5 available in GitHub Copilot

Claude Haiku 5.5 is available on Copilot Pro, Pro+, Max, Business and Enterprise at provider list price.

Copilot code review now defaults to Balanced effort

The Default review effort level for Copilot code review now uses Balanced for new and existing repositories and organizations, effective September 28, 2026. Settings explicitly set to Lite were kept.

Who is affected:

Teams using Copilot code review on the Default effort setting.

What to do:

Set the review effort to Lite if you want the previous lighter reviews.

Copilot removes Claude Opus 4.7, Kimi K2.7 Code, Gemini 3.5 Flash and 3.6 Flash

GitHub removed four models from all Copilot experiences, announced on September 3. Suggested replacements are Claude Opus 5.5, Kimi K3 and Gemini 3.8 Flash.

Who is affected:

Copilot users and workflows pinned to any of the four models.

Action needed:

Switch to the suggested replacement models.

GPT-6.1 Sol available in GitHub Copilot

GPT-6.1 Sol is available on Copilot Pro+, Max, Business and Enterprise at provider list price.

Claude Sonnet 5.5 available in GitHub Copilot

Claude Sonnet 5.5 is available on Copilot Pro, Pro+, Max, Business and Enterprise at provider list price.

Claude Opus 5.5 available in GitHub Copilot

Claude Opus 5.5 is available on Copilot Pro+, Max, Business and Enterprise at provider list price. GitHub says it resolved tasks about as well as Opus 5 with significantly fewer steps and tokens.

GPT-6 Sol and GPT-6 Luna available in GitHub Copilot

GPT-6 Sol is available on Copilot Pro+, Max, Business and Enterprise, and GPT-6 Luna also on Pro, under usage-based billing.

Grok 4.7 available in GitHub Copilot

Grok 4.7 rolled out to Copilot Pro, Pro+, Max, Business and Enterprise, enabled automatically under default model enablement.

Copilot auto model selection gets efficiency, balance and intelligence tiers

Copilot auto model selection now offers three tiers that weigh cost, quality and response time differently. Usage is charged for the model auto picks, with a 10% discount for paid subscribers.

Who is affected:

Copilot users who rely on auto model selection.

What to do:

Pick the tier that matches your cost and quality needs.

GPT-6 Astra available in GitHub Copilot

GPT-6 Astra is available on Copilot Pro+, Max, Business and Enterprise at provider list price, and is enabled automatically under default model enablement.

Copilot Business and Enterprise sign-ups reopen for card and PayPal customers

GitHub is gradually reopening Copilot Business and Enterprise sign-ups for customers paying by card or PayPal, with stricter account vetting. New seats must be paid before access, and usage beyond the included amount may need extra payment.

Gemini 3.8 Flash available in GitHub Copilot at introductory pricing

Gemini 3.8 Flash is available on Copilot Pro, Pro+, Max, Business and Enterprise, billed at introductory provider pricing through December 31, 2026. It is enabled automatically under default model enablement.

Claude Fable 5.1 available in GitHub Copilot, with data retention by default

Claude Fable 5.1 is available on Copilot Pro+, Max, Business and Enterprise. Unlike other Claude models in Copilot it requires data retention so Anthropic can run safety classifiers. Eligible enterprises can use zero data retention through the end of 2026.

Who is affected:

Copilot organizations with zero data retention requirements.

What to do:

Copilot admins: review the data retention terms before enabling the Fable 5.1 policy.

Copilot removes Gemini 3.1 Pro and older Claude models

On September 1, 2026 Copilot deprecated Gemini 3.1 Pro (replace with Gemini 3.7 Flash), Claude Opus 4.5 and 4.6, Claude Sonnet 4.5 and 4.6, and Raptor Mini.

Who is affected:

Copilot users and integrations pinned to those models.

Action needed:

Switch to the suggested alternatives. Admins may need to enable them.

Copilot model access now follows the organization that pays for your usage

For users with Copilot seats in more than one organization, model access is now decided only by the billing organization. Before, a model was usable if any of the organizations enabled it. Access that comes entirely through an enterprise is unchanged.

Who is affected:

Copilot users with seats in several GitHub organizations.

What to do:

Check which organization is listed under Usage billed to if a model disappeared.

Copilot Business and Enterprise seats billed upfront for card and PayPal customers

New Copilot Business and Enterprise seats require payment before access. From October 1, 2026, existing card and PayPal customers are charged upfront for assigned seats at the start of each billing cycle. Prices are unchanged.

Who is affected:

Copilot Business and Enterprise customers paying by credit card or PayPal.

What to do:

Check seat assignments before your next billing cycle starts.

Grok 4.6 available in GitHub Copilot

Grok 4.6 rolled out to Copilot Pro, Pro+, Max, Business and Enterprise at provider list price. Business and Enterprise admins must turn on the policy, which is off by default.

Gemini 3.7 Flash available in GitHub Copilot

Gemini 3.7 Flash rolled out to Copilot Pro, Pro+, Max, Business and Enterprise at provider list price.

MAI-Code-1.1-Flash replaces MAI-Code-1-Flash in Copilot at a 73% lower price

Microsoft MAI-Code-1.1-Flash rolled out in Copilot with vision support and a list price 73% below MAI-Code-1-Flash. Free and Student users get it through auto model selection. MAI-Code-1-Flash was removed on September 10, 2026.

Who is affected:

Copilot users and workflows using MAI-Code-1-Flash or auto model selection.

What to do:

Switch any pinned MAI-Code-1-Flash workflows to MAI-Code-1.1-Flash.

Kimi K3 available in GitHub Copilot

Kimi K3, hosted by GitHub on Fireworks AI, rolled out to Copilot Pro, Pro+, Max, Business and Enterprise at $3 per 1M input tokens, $15 per 1M output tokens and $0.30 per 1M cached input tokens.

Who is affected:

Copilot users looking for a cheaper frontier-level coding model.

What to do:

Business and Enterprise admins must enable the Kimi K3 policy.

Copilot removes Gemini 2.5 Pro and Gemini 3 Flash

GitHub removed Gemini 2.5 Pro and Gemini 3 Flash from all Copilot experiences, announced on July 2. Suggested replacements are Gemini 3.1 Pro (Preview) and Gemini 3.6 Flash.

Who is affected:

Copilot users and workflows pinned to Gemini 2.5 Pro or Gemini 3 Flash.

Action needed:

Switch to Gemini 3.6 Flash or another supported model.

New Copilot models turn on by default for Business and Enterprise from August 26

GitHub added a default enablement policy so newly GA models are on by default for Business and Enterprise from August 26, 2026, unless admins opt out. Open-weight models and models outside GitHub's data retention agreement are excluded.

Who is affected:

Copilot Business and Enterprise admins who approve models manually.

What to do:

Set the policy to disabled before August 26 if you want to approve each model.

Grok 4.5 available in GitHub Copilot

Grok 4.5 rolled out to Copilot Pro, Pro+, Max, Business and Enterprise, billed at provider list pricing. The policy is off by default for organizations.

Who is affected:

Copilot users who want xAI models.

What to do:

Business and Enterprise admins must enable the Grok 4.5 policy.

Claude Opus 5 available in GitHub Copilot

Claude Opus 5 is available on Copilot Pro+, Max, Business and Enterprise at provider API list price. GitHub notes its cyber safeguards may block some security-related requests.

Who is affected:

GitHub Copilot users on Pro+ and above.

What to do:

Copilot Business and Enterprise admins: enable the Claude Opus 5 policy if you want it.

GPT-5.6 Sol, Terra and Luna available in GitHub Copilot

GPT-5.6 Sol is available on Copilot Pro+, Max, Business and Enterprise. Terra and Luna are also available on Pro. All are billed at provider list price, and Business and Enterprise admins must turn on the policy, which is off by default.

Who is affected:

GitHub Copilot users choosing OpenAI models.

What to do:

Copilot Business and Enterprise admins: enable the GPT-5.6 policy if you want these models.

Kimi K2.7 Code is the first open-weight model in Copilot

Kimi K2.7 Code, hosted by GitHub on Microsoft Azure, became selectable in Copilot, starting with Pro, Pro+ and Max, billed at provider list pricing.

Who is affected:

Copilot users looking for a lower-cost model.

What to do:

Business and Enterprise admins must enable the policy, which is off by default.

Copilot Free and Student plans lose manual model choice

Copilot Free and Student now use auto model selection as the default and only model selection experience.

Copilot individual plan sign-ups reopen

GitHub began reopening sign-ups for Copilot Student, Pro, Pro+ and Max gradually, and added pay-as-you-go for usage beyond plan limits.

Copilot switches all plans to usage-based billing

All Copilot plans now bill by GitHub AI Credits consumed, with monthly included usage and optional extra spend. Copilot code review also consumes GitHub Actions minutes, and a new Copilot Max plan is available as an upgrade.

Who is affected:

Every Copilot subscriber and organization.

Action needed:

Set spending budgets and check what your plan includes.

Gemini 3.5 Flash in GitHub Copilot at a 14x multiplier

Gemini 3.5 Flash rolled out to Copilot Pro, Pro+, Business and Enterprise with a tentative 14x premium request multiplier.

Who is affected:

Copilot users choosing between models by cost.

What to do:

Business and Enterprise admins must enable the Gemini 3.5 Flash policy.

GPT-5.3-Codex becomes the Copilot base model for Business and Enterprise

GPT-5.3-Codex replaced GPT-4.1 as the base model for all Copilot Business and Enterprise organizations, at a 1x premium request multiplier. It is a long-term support model available through February 4, 2027.

Who is affected:

Copilot Business and Enterprise organizations that have not approved other models.

What to do:

Review premium request usage, since the base model now has a 1x multiplier.

GitHub Copilot moves all plans to usage-based AI Credits from June 1, 2026

GitHub announced that every Copilot plan would switch from premium requests to a monthly allotment of GitHub AI Credits on June 1, 2026, with usage priced on input, output and cached tokens at each model's API rates. The Hacker News thread drew 767 points and 553 comments.

Who is affected:

All GitHub Copilot subscribers, especially heavy agent users.

Action needed:

Review your AI Credits usage in Copilot settings and set a budget.

Copilot pauses Pro sign-ups, tightens limits and removes Opus from Pro

GitHub paused new sign-ups for Copilot Student, Pro and Pro+, tightened usage limits on individual plans, and removed Opus models from Copilot Pro. Subscribers could cancel for a refund until May 20.

Who is affected:

Copilot Pro, Pro+ and Student subscribers.

Action needed:

If Opus on Pro was important to you, upgrade to Pro+ or request a refund before May 20.

Copilot enforces new usage limits and retires Opus 4.6 Fast from Pro+

GitHub began rolling out service reliability limits and per-model limits for Copilot, and retired Opus 4.6 Fast for Copilot Pro+ users.

Who is affected:

Heavy Copilot users who send many concurrent requests.

Action needed:

Spread requests over time, switch models or use Auto when you hit a model limit.