GitHub Copilot
The widest model catalogue and the deepest agent surface in any coding tool — but since 1 June 2026 the $10 sticker price is only the entry fee on a token meter, and heavy agent users are burning a month of credits in a week.
PRICING
| Free | $0 | 2,000 completions/month, limited chat and agent usage, auto model selection only |
| Pro | $10/mo | unlimited completions, cloud agent, model selection, $15/mo in AI credits |
| Pro+ | $39/mo | premium models incl. Opus, audit logs, $70/mo in AI credits |
| Max | $100/mo | priority access to new models, $200/mo in AI credits |
| Business | $19/seat/mo | pooled credits, centralized policy control, content exclusion |
| Enterprise | $39/seat/mo | larger pooled credits, codebase indexing, enterprise policy inheritance |
Usage-based billing went live for all plans on 1 June 2026, replacing premium requests and model multipliers. Chat, agent mode, code review, the cloud agent and the CLI all draw AI credits at per-token model rates; code review additionally consumes GitHub Actions minutes. Auto model selection carries a discount and some baseline models draw no credits. New self-serve Copilot Business sign-ups for organisations on GitHub Free and Team have been paused since 22 April 2026.
checked 2026-08-07 · vendor pricing page
Element scores
Strengths
Model access is the widest of any coding tool — OpenAI, Anthropic, Google, xAI, Moonshot and Microsoft in one picker, with a published retirement calendar so you know when a model disappears. The agent surface is equally broad: agent mode GA across VS Code, Visual Studio, JetBrains, Eclipse and Xcode; a cloud agent that turns an assigned issue into a reviewed pull request; a CLI; and a desktop app (Build 2026) running parallel sessions in isolated git worktrees with Interactive, Plan and Autopilot modes. Customisation is version-controlled and reviewable: instructions files, Agent Skills, .agent.md custom agents, Copilot Spaces. Governance stacks properly across enterprise, organisation and repository, with content exclusion, the public-code duplication filter, a usage-metrics REST API and a 180-day audit log. The cheap-model ladder (GPT-5 mini, nano, Haiku 4.5, Flash, Raptor mini) plus a discounted Auto mode is a real cost lever.
Honest dings
The 1 June 2026 move to AI Credits is the headline risk. Subscription prices did not change but the meter underneath did, and the community response was blunt: one Pro+ subscriber documented burning 15% of a monthly allowance in a single 2–3 hour session on an unchanged workflow, and a Max subscriber reported spending another $100+ on top of the $100 plan. Budgeting agent work is now a modelling exercise, not a line item. Content exclusion — the control most enterprises lean on — does not apply in Copilot CLI or IDE agent mode per GitHub's own documentation, which is precisely where the agent touches most files. The audit log proves policy configuration, not code provenance, so it cannot answer the question auditors actually ask. Customer fine-tuning was retired with no replacement. And there is no evaluation surface: you can review the agent's code, but you cannot systematically test the agent.
Sources
Every audit lists the research it rests on — transparency and traceability are the product. Tools evolve: each audit is a snapshot of its audit date, and re-audits supersede older versions (kept below for reference).
- github.com/features/copilot/plans — Official plans and pricing: Free/Pro $10/Pro+ $39/Max $100, AI credit allowances ($15/$70/$200), base credits and flex allotment, full per-plan model catalogue, agent mode and cloud agent availability (accessed 2026-08-07)
- docs.github.com/en/copilot/get-started/plans — Official plan comparison incl. Business $19 and Enterprise $39 per granted seat, Copilot Student, and the 22/04/2026 pause on new self-serve Copilot Business sign-ups (accessed 2026-08-07)
- docs.github.com/en/copilot/get-started/features — Official feature inventory: Copilot CLI, desktop app, cloud agent, third-party coding agents (preview), code review, agent mode, Spaces, custom instructions, Copilot Memory (preview), MCP servers, Agent Skills, custom agents (accessed 2026-08-07)
- docs.github.com/en/copilot/reference/ai-models/supp… — Official model catalogue per plan (GPT-5.6 Sol/Terra/Luna, Claude Opus 5, Fable 5, Gemini 3.6 Flash, Grok 4.5, Kimi K2.7 Code, MAI-Code-1-Flash, Raptor mini) and the model retirement schedule (accessed 2026-08-07)
- docs.github.com/en/copilot/reference/copilot-billin… — Official per-token pricing and the AI-credit mechanic (1 AI credit = $0.01); GPT-5 mini $0.25/M input vs GPT-5.5 $5.00/M input evidences the cost ladder (accessed 2026-08-07)
- docs.github.com/en/copilot/concepts/agents/github-c… — Official: parallel agent sessions in isolated git worktrees, Interactive/Plan/Autopilot modes, BYOK model selection, reasoning-effort control, scheduled automations, cloud sandbox sessions (accessed 2026-08-07)
- docs.github.com/en/copilot/how-tos/configure-conten… — Official content-exclusion configuration at repo/org/enterprise level — and the stated limitation that Copilot CLI and IDE agent mode do not support content exclusion (accessed 2026-08-07)
- docs.github.com/en/copilot/how-tos/administer-copil… — Official governance layers: enterprise-over-org policy inheritance, MCP server policy, per-partner toggles for Anthropic Claude and OpenAI Codex agents (accessed 2026-08-07)
- mlq.ai/earnings/highlight/MSFT-github-copilot-reach… — Satya Nadella on the Microsoft Q1 FY2026 earnings call: GitHub Copilot 'now with over 26 million users' — official user figure (accessed 2026-08-07)
- github.com/orgs/community/discussions/197735 — Official GitHub community thread with admin response: Pro+ subscriber consumed ~15% of the monthly allowance in one 2–3h session after the 1 June 2026 change; a Max subscriber reports $100+ of extra credits monthly (01/08/2026) (accessed 2026-08-07)
- github.com/orgs/community/discussions/161278 — Official GitHub community discussion confirming customer fine-tuning of Copilot models is deprecated for Business and Enterprise, with no reintroduction announced (accessed 2026-08-07)
- re-entry.ai/blog/github-copilot-enterprise-audit-lo… — Independent analysis (13/07/2026): Copilot Enterprise audit logs capture policy toggles, seat assignments and exclusion changes but not suggestion acceptance or code provenance (accessed 2026-08-07)
- anomity.ai/blog/governing-github-copilot-in-the-ent… — Independent 2026 enterprise governance guide: three stacked policy layers, duplication-detection filter posture, audit-log action filters, 180-day retention, documented exclusion gaps (accessed 2026-08-07)
- digitalapplied.com/blog/github-copilot-app-agent-na… — Independent review (06/06/2026) of the Copilot desktop app announced at Microsoft Build on 02/06/2026, and the Copilot SDK reaching GA across six languages (accessed 2026-08-07)
- faros.ai/blog/best-ai-coding-agents-2026 — Independent 2026 comparative review positioning Copilot as the 'pragmatic default' and recording power-user criticism on reasoning depth, quotas and opaque model choices (accessed 2026-08-07)
- tech-insider.org/au/github-copilot-usage-based-bill… — Independent coverage of the 01/06/2026 usage-based billing switch: $0.04 per additional premium request, developer backlash, 0x-multiplier baseline models as the cost escape valve (accessed 2026-08-07)