Text of about.gitlab.com › compare › gitlab-vs-github as Styx crawled it on Nov 18, 2026. The page may have changed since. Only its words are kept, without pictures or layout, and only the first part of a long page.

Open the live page

GitLab vs GitHub

To search repositories and projects, login to gitlab.com . Suggestions GitLab Duo Agent Platform Code Suggestions (AI) CI/CD GitLab on AWS GitLab on Google Cloud Why GitLab? Offer: Migrating from GitHub? Get 12 months of GitLab free GitLab vs. GitHub The AI era demands more from your software delivery stack Get free trial See how we compare What are the fundamental differences between GitLab and GitHub? GitHub GitLab GitHub Add-ons required, gaps remain GitHub’s security scanning and advanced CI/CD are sold as separate paid add-ons on top of the base platform price.

Copilot is now included in GitHub Enterprise, but every AI request is metered against an enterprise budget from first use, with no credits included per seat. Value stream analytics, DORA metrics, and DAST are not available from GitHub at any tier and require third-party tools to fill. GitLab is a unified platform GitLab puts planning, source code, CI/CD, security, and deployment into a single application with one permission model, one audit trail, and shared analytics.

That means fewer tools to run, fewer integrations to maintain, and one place to apply policies and AI across the whole lifecycle. GitHub Context with code- and repo-centric AI GitHub has added event-triggered Copilot automations, GitHub Agentic Workflows and agentic autofix for code scanning alerts. These are separate products built on repositories and GitHub Actions, rather than agents working from one data model across planning, CI/CD, and security.

Full-lifecycle, policy-aware AI context GitLab Duo Agent Platform runs on a single DevSecOps platform, so agents see issues, epics, merge requests, pipelines, environments, and security findings in one data model. That unified SDLC and policy context lets GitLab Duo power agents for planning, CI/CD troubleshooting, and security remediation without stitching together separate products or backends. GitLab Orbit (GA), available as an add-on for Premium and Ultimate, connects code, merge requests, work items, pipelines, and vulnerabilities in a single, continuously updated view, and makes that context available to agents on GitLab Duo Agent Platform and external agents through the GitLab MCP server.

GitHub AI centered on code and repository tasks GitHub Copilot and the GA Copilot coding agent automate code- and repo-scoped tasks like implementing features, fixing bugs, improving tests, and updating documentation via pull requests and GitHub Actions. For broader SDLC automation across cross-project planning, CI/CD orchestration, and security remediation, assembly of multiple tools and workflows is required, while GitLab focuses on SDLC-wide agents and flows on one platform.

GitLab’s AI works across your entire delivery chain GitLab Duo Agent Platform provides agents and flows that automate planning, coding, CI/CD, security, documentation, onboarding, and modernization. These agents help break down epics into issues, convert and optimize pipelines, diagnose and fix failed builds, triage and remediate vulnerabilities, generate architecture docs and onboarding guides, and help modernize codebases with custom flows on a single platform.

GitHub GitHub trains AI models by default for some tiers As of April 24, 2026, GitHub uses Copilot interaction data — inputs, outputs, code snippets, and associated context — to train AI models by default for Free, Pro, and Pro+ users. Users must manually opt out. This data may also be shared with GitHub affiliates, including Microsoft for AI development purposes. Copilot Business and Copilot Enterprise customers are exempt from this model training under their data protection agreements.

GitLab doesn’t use customer code to train models. Period. GitLab doesn’t train its models on any customer code at any tier, including free, Premium, and Ultimate. This means there’s no way any contractors, open source contributors, or developers on individual-tier accounts could have code entering training pipelines without company visibility or control. GitHub Tied to Microsoft and Azure GitHub is owned by Microsoft, and the GitHub Copilot SDK is becoming Microsoft Copilot SDK.

Copilot requests to GitHub-hosted models pass through a GitHub proxy hosted in Azure. GitHub Actions now runs some Linux jobs on AWS as well as Azure, but the Copilot proxy remains in Azure. When GitHub Enterprise Cloud is purchased through a Microsoft Enterprise Agreement, a connected Azure subscription is needed to use Copilot, GitHub Advanced Security, or usage beyond included amounts. GitLab is cloud agnostic GitLab is not commercially tied to any cloud provider.

GitHub Broad cloud model catalog and BYOM GitHub Copilot offers a wide range of cloud-hosted models (GPT-6 families, Claude, Gemini, Grok, and more) plus “bring your own API key” options for several providers. Admins can control which models are allowed at the enterprise, organization, or team level, and Copilot Fusion can pick a model automatically based on an effort level. Flexible model choice on one platform GitLab Duo Agent Platform lets you use GitLab-managed models, open-weight models, or your own self-hosted deployments (Anthropic, OpenAI, Mistral, Llama, and, in beta, any OpenAI-compatible endpoint) through the AI Gateway.

Admins choose a model for each feature at the instance or top-level group, and can restrict Agentic Chat to approved models. This gives you model flexibility without fragmenting your DevSecOps platform. GitHub Credits tied to licenses purchased Copilot chat, agents, and code review bill in AI credits ($0.01 each), converted from per-model token rates. Code completions remain unmetered. Copilot Business ($19) and Copilot Enterprise ($39) seats include 1,900 and 3,900 credits per user.

Copilot included with GitHub Enterprise and the $10 Copilot Standalone include no credits, so usage is metered from the first request. Reallocatable annual commitment GitLab Duo Agent Platform usage is billed in GitLab Credits, with Code Suggestions, Code Review Flow, and SAST false positive detection and vulnerability resolution (Ultimate) priced per execution regardless of model. GitLab Flex lets you pre-commit to an annual contract balance and reallocate it across seats and credits.

Customers can shift spend fluidly between seats and AI credits month to month without renegotiating the contract. GitHub Cloud-first AI execution GitHub Copilot runs in provider clouds (Azure/OpenAI, Anthropic, Google, xAI), even when you supply your own API keys. Two exceptions are Copilot Fusion which can route some tasks to models on a developer’s machine (public preview), and Copilot CLI can run air-gapped on GitHub Enterprise Server with an admin-configured model provider (technical preview).

The coding agent and agentic code review run only in GitHub’s cloud and aren’t available to Enterprise Server customers. Self-hosted models and agents, including air-gapped GitLab Duo Self-Hosted and GitLab Duo Agent Platform Self-Hosted let you run models and agents on your own infrastructure (on-prem or private cloud) via a self-hosted AI Gateway, including fully offline and air-gapped setups. You can also mix self-hosted and GitLab-managed models in a hybrid configuration when internet access is allowed.

On GitLab Dedicated (Ultimate), a GitLab-hosted AI Gateway in your instance’s Amazon Web Services (AWS) region routes the features you choose to their own model endpoints. GitHub Security requires paid add-ons GitHub Advanced Security adds CodeQL-based code scanning, secret scanning, and dependency review on top of GitHub, purchased separately from the base plans. It does not provide native DAST or API security testing at any tier and relies on marketplace actions and partner tools for container image scanning.

Built-in security GitLab’s SAST, secret detection, and container scanning are included in every GitLab tier, including Free and Premium, with no separate paid add-ons required. Scans run automatically on every m

The rest of this page is on the live site.