GitHub

Executive Summary

What it is: GitHub Copilot spans every major IDE, CLI, github.com, and its desktop app, from $0 (Free) to $100/mo Max with $200 total AI credits, Business at $19 and Enterprise at $39 per user/month. It now exposes 31 models from 8 providers including Claude Opus 5, GPT-5.6 Sol/Terra/Luna, Grok 4.6, Gemini 3.7 Flash, Kimi K3, and Microsoft's MAI-Code-1.1-Flash. Source: https://github.com/features/copilot/plans

What to watch out for: Business/Enterprise signups were closed and reopen Sep 1 with upfront seat billing from Oct 1, while the post-promo included-usage amount after the $30/$70 credits lapse is still undisclosed. Chat data retention moves from 28 days to account lifetime from ~Sep 28, code review defaults jump from Lite to Balanced effort (more credits per review), and six models were deprecated Sep 1. Source: https://github.blog/changelog/2026-08-28-upcoming-changes-to-github-copilot-policies-and-billing/

Bottom line: Copilot remains the broadest multi-model surface and the cheapest managed entry to frontier models ($19 Business with IP indemnity), but its billing model is in flux: enterprises should pin model policies via the new Global model policy and quantify the Balanced-review cost change before Sep 28. Source: https://github.blog/changelog/2026-08-26-global-model-policy-generally-available

Key Terms

Latest Changes

Changes since the 2026-07 report.

July watch items verified:

  • Business/Enterprise promo credits ($30/$70) expiring after August: superseded by a billing model change. On August 28 GitHub announced that Business/Enterprise seat prices are "not changing" ($19/$39 per user/month), that new credit-card/PayPal signups reopen September 1 after a closure period, and that from October 1 all assigned seats incur an upfront charge with included usage possibly prorated monthly. Neither the announcement nor the plans page states the ongoing included-usage amount after the promo credits lapse; this remains undisclosed. Sources: GitHub – 2026 08 28 Upcoming Changes To Github Copilot Policies And Billing , GitHub – Plans
  • Default model enablement effective August 26: confirmed. The Global model policy went GA the same day, giving admins one place to manage model access. Source: GitHub – 2026 08 26 Global Model Policy Generally Available
  • UK CMA probe over Copilot-linked price hikes: still-pending. No public update found in August.

New August changes:

Plans

Individual Plans

Plan Price Total AI credits/mo Base + Flex Key inclusions
Free $0 Limited chat and agent usage n/a 2,000 inline completions/mo, limited chat/agent; model list restricted to Claude Haiku 4.5, OpenAI GPT-5 mini, and Raptor mini (Preview); agent mode + CLI included
Pro $10/mo $15 $10 + $5 flex Unlimited completions, all mainline models (Sonnet 5, Opus 4.7/4.8/5, GPT-5.6 Terra/Luna, Gemini 3.5-3.7 Flash, Grok 4.5/4.6, Kimi K3, MAI-Code-1.1-Flash); code review; cloud agent
Pro+ $39/mo $70 $39 + $31 flex Everything in Pro + Fable 5, GPT-5.5, GPT-5.6 Sol; third-party agent delegation (Claude, Codex)
Max $100/mo $200 $100 + $100 flex Everything in Pro+ + full model matrix

Source (model matrix verified in-browser September 2): GitHub – Plans

Business / Enterprise

Plan Price Included usage Notes
Business $19/user/mo (unchanged) $30/mo promo credits were scheduled to expire after August; post-promo amount undisclosed New signups (card/PayPal) reopen Sep 1; upfront seat billing from Oct 1
Enterprise $39/user/mo (unchanged) $70/mo promo credits were scheduled to expire after August; post-promo amount undisclosed Adds codebase indexing, custom models, github.com integration

Business/Enterprise include IP indemnity, SAML SSO, user management, usage metrics, enterprise-grade security, and no training on customer data. Source: GitHub – 2026 08 28 Upcoming Changes To Github Copilot Policies And Billing

Overage: Additional AI credits purchasable at $0.01/credit with budgets and alerts at 75/90/100%; admins control whether paid overage is allowed on Business/Enterprise. Source: GitHub – Plans

API Pricing

GitHub Copilot is not API-first: there is no per-token API pricing for Copilot itself. Cost is metered via AI Credits (1 credit = $0.01) consumed by model choice and task complexity. Third-party models reachable through Copilot price at their own API rates elsewhere in this report. Code completions and next-edit suggestions do not consume credits on paid plans. Source: GitHub – Plans

Model Performance / Benchmarks

No new GitHub-published Copilot-specific benchmarks this month. Relevant incoming-model scores are reported in the suppliers' own sections (Anthropic Fable 5.1, Google Gemini 3.7 Flash, xAI Grok 4.6, Moonshot Kimi K3). No published coding benchmarks exist for MAI-Code-1.1-Flash. Source: GitHub – Changelog

Latest News

Business Model Reset: Signups Reopen, Upfront Billing (Aug 28, 2026)

GitHub will "reenable sign-ups for new Copilot Business and Copilot Enterprise customers paying by credit card or PayPal" starting September 1, confirming new-customer signups had been closed. Billing changes follow: every assigned seat incurs an upfront charge at the start of each cycle (existing card/PayPal customers from October 1), included usage may be prorated across the month, and fully canceling and returning may trigger updated vetting. Prices stay $19/$39. Source: GitHub – 2026 08 28 Upcoming Changes To Github Copilot Policies And Billing

Six-Model Purge and Lineup Refresh (Aug 31 / Sep 1, 2026)

Gemini 3.1 Pro, Claude Opus 4.5/4.6, Sonnet 4.5/4.6, and Raptor Mini were deprecated September 1, while August added Kimi K3, MAI-Code-1.1-Flash, Gemini 3.7 Flash, and Grok 4.6, and Fable 5.1 arrived September 1. The lineup now spans 31 models across 8 providers (Anthropic, OpenAI, Microsoft, Google, xAI, Moonshot, GitHub Raptor, plus third-party agent delegation). Sources: GitHub – 2026 08 31 Selected Github Copilot Models Deprecated , GitHub – Plans

Copilot Everywhere: Slack, Teams, JetBrains Harness, Agent Plugins (August 2026)

A new Copilot experience shipped in Slack (Aug 21) and shared agentic work in Microsoft Teams (Aug 21); the Copilot harness went GA in JetBrains (Aug 24); Agent Plugins 1.0 landed across VS Code, CLI, and the Copilot app (Aug 12); Copilot memory and Ollama (local model) support arrived in JetBrains (Aug 11). Sources: GitHub – 2026 08 21 The New Github Copilot Experience In Slack , GitHub – 2026 08 11 Copilot Memory And Ollama In Github Copilot For Jetbrains

Security Research: Copilot Cowork Sandbox Bypass (Aug 25-26, 2026)

PromptArmor published "Bypassing Copilot Cowork's Sandbox" describing attacker control of Microsoft Copilot Cowork with sandbox escape (Microsoft's sibling product to GitHub Copilot). HN: News – Item . Separately, GitHub denied that a bug found by Wiz's AI agent was written by Copilot Autofix. Ai-News – Github Denies Copilot Autofix Wrote The Bug Wizs Ai Agent Found

Community Signals

Billing and Cost-Creep Concerns

whirlwin posted "GitHub Copilot reviews cost creep" (3 points) linking the August 28 policy/billing announcement, reflecting concern that the Balanced-by-default review effort and unified experience will raise credit consumption. News – Item . The code-review-can-approve-PRs change was also submitted to HN. News – Item

Autocomplete vs Agent Usage

jawuilp's "Ask HN: Are you still using AI code autocomplete?" (5 points, 16 comments, Aug 26) asked whether developers still use Cursor Tab or Copilot autocomplete now that agentic assistants dominate, noting "By the time I started having access to decent AI autocomplete tools, things like Cline and Augment were already around." News – Item

Adoption Data

JetBrains' Developer Ecosystem Survey 2026 (published Aug 18) provides third-party adoption estimates for Claude Code, Codex, Cursor, JetBrains Junie, and other agents. Source: Jetbrains – Ai Coding Agent Adoption 2026

Enterprise Readiness

Feature Available? Details
SSO (SAML) Yes Business and Enterprise plans. Source: GitHub – Plans
SSO (OIDC) No Not listed on plans pages.
SCIM Partial User management via org/enterprise APIs; full SCIM provisioning documented for Enterprise through GitHub's identity management. Source: GitHub – Copilot
Audit logs Yes Enterprise; usage metrics and audit log tracking expanded through 2026. Source: GitHub – Plans
IP indemnity Yes Business and Enterprise (Microsoft Copilot Copyright Commitment). Source: GitHub – Plans
Data residency Partial Enterprise-managed settings, IP allowlists, MCP allowlists (Aug 6); no region-pinned inference option published. Source: GitHub – 2026 08 06 Mcp Allowlists In Enterprise Managed Settings
HIPAA Partial Covered via GitHub's Enterprise compliance programs (BAA available); not itemized on Copilot plans page. Source: GitHub – Trust Center
Air-gapped / on-prem Partial GitHub Enterprise Server 3.22 RC (Aug 11) supports Copilot; full on-prem Copilot feature parity is not documented. Source: GitHub – 2026 08 11 Github Enterprise Server 3 22 Release Candidate
SLA No No Copilot-specific SLA published.
Admin controls (RBAC) Yes Global model policy (Aug 26), enterprise managed settings across IDEs/CLI/app, MCP allowlists, budget controls with expiration dates (Sep 1), per-user budgets. Sources: GitHub – 2026 08 26 Global Model Policy Generally Available , GitHub – 2026 09 01 Set An Expiration Date For Individual User Budgets

Transparency Gaps

Gap Details Severity
Post-promo Business/Enterprise included usage The $30/$70 promo credits were scheduled to end after August. GitHub's August 28 billing announcement and the plans page do not state the ongoing included amount; the new upfront-billing model makes monthly cost projection harder, not easier. High
Closed-signup period Business/Enterprise signups were closed for an undisclosed period before the Sep 1 reopening; GitHub never published how long or why beyond "availability and reliability". Medium
Credit burn rates Per-model credit consumption for typical sessions still has no published estimates, though the new per-model token breakdown (Aug 11) improves after-the-fact visibility. High
Unified experience cost impact Moving review defaults from Lite to Balanced (Sep 28) raises per-review credit consumption by an undisclosed amount. Medium
Chat retention change Lifetime retention of chat data (from ~Sep 28) is a material privacy shift for enterprises; the retention period for agent-session artifacts vs chat is not fully specified. Medium
GPT-5.2 family pricing GPT-5.2, GPT-5.2-Codex, GPT-5.3-Codex, and GPT-5.5 still lack published per-interaction credit costs distinct from the model matrix. Medium
Flex allotment instability Flex credits remain "variable" and subject to change without notice. Medium
MAI-Code-1.1-Flash specs No published benchmarks, context window, or credit multiplier for Microsoft's in-house coding model. Low