- GitHub Copilot has surpassed 20 million users, with roughly 4.7 million on paid plans, making it the most widely adopted AI coding assistant.
- Recent updates include new slash commands in the Copilot app, a standalone CLI, and enhanced agent mode capabilities in VS Code.
- Kimi K3, an open-weight model, is now generally available in Copilot, with gradual rollout and admin controls for Business and Enterprise plans.
- Microsoft is internally managing AI token usage, shifting to GPT-5.6 as default and setting budget targets, reflecting broader industry trends.
GitHub Copilot has cemented its position as the most widely used AI coding assistant, crossing the 20 million user mark in 2025. According to a 2026 developer-tools roundup, about 4.7 million of those users are on paid subscriptions. This scale gives Copilot a significant edge, even as competitors like Cursor and Claude Code win specific benchmarks. The tool’s deep integration with GitHub and Visual Studio Code makes it the default choice for many developers, and recent updates have only strengthened its appeal.
The product has evolved far beyond simple autocomplete. It now bundles inline completions, a conversational chat panel, multi-file editing, and an autonomous agent mode under one subscription. This shift means setting up and using Copilot involves configuring several distinct tools, not just installing a smarter autocomplete. The latest updates across the desktop app, CLI, and VS Code focus on helping users resume work, organize sessions, and ask questions without losing context.
New Features and Workflow Enhancements
The GitHub Copilot app has received several notable updates. The Auto feature now shows which model handled each request, along with AI credit and cache details when available. Users can jump directly into shared sessions or use the /side command to explore parallel questions without disrupting their main task. The updated Impeccable skill applies to mobile interfaces, and sessions now start and switch more efficiently.
The Copilot CLI has also seen significant improvements. A new Sessions sidebar allows easy management of multiple concurrent sessions, with shortcuts to open, close, and navigate between them. An experimental /worktree command creates an isolated workspace for exploring changes without disrupting current work. The /rewind command can restore conversations and files without Git, and live tool-call durations help identify slow commands.
VS Code 1.132 brings several enhancements to the Copilot experience. The integrated browser now supports element-level feedback, allowing users to select and annotate specific parts of a web page. Dictation uses a multilingual on-device model by default, keeping audio local. The /btw command opens a side chat without interrupting the agent’s current turn, and Markdown diffs can now be reviewed in a hybrid editor.
Pricing and Model Availability
GitHub restructured Copilot’s pricing around an AI credits system in June 2026. One credit equals $0.01, and code completions remain unlimited on all paid plans. The Free tier offers 2,000 completions and 50 chat requests per month, while Pro costs $10/month with $15 in AI credits. Pro+ at $39/month includes $70 in credits and premium models, and Business and Enterprise plans offer org-pooled credits and priority access.
Kimi K3, an open-weight model, is now generally available in GitHub Copilot. The model shows frontier-level abilities on agentic coding with cost-effective pricing, hosted by GitHub on Fireworks AI. It is rolling out gradually to Copilot Pro, Pro+, Max, Business, and Enterprise plans, and can be selected in the model picker across VS Code, Visual Studio, Copilot CLI, and other platforms. For Business and Enterprise, admins must enable the Kimi K3 policy before anyone can use it.
Microsoft is also managing its internal AI usage more carefully. An internal memo from executive vice president Jay Parikh emphasized optimizing for “more impact per token” rather than just consuming more. OpenAI’s GPT-5.6 is now the default model for internal use, and each division has an AI token budget target. This reflects a broader industry trend, with companies like Amazon, Adobe, and Meta implementing similar measures.
Competitive Landscape and Market Position
GitHub Copilot’s main advantage is distribution. It ships as a first-party extension inside the editor most professional developers already use, and bills through the same GitHub account that hosts their repositories. This low switching cost is a significant factor in its adoption. In contrast, Cursor and Windsurf (now Devin Desktop) are standalone editors built on VS Code forks, requiring users to switch editors entirely.
Pricing is a key differentiator. Copilot’s Pro plan at $10/month is significantly cheaper than Cursor’s $20/month or Devin Desktop’s $20/month. For teams, Copilot Business at $19/seat/month is roughly half the cost of the other two tools’ team tiers. This per-seat economics makes Copilot a more attractive option for larger organizations, even if competitors offer more advanced agent features.
Independent reviews consistently rank Cursor’s Composer as best-in-class for multi-file editing. However, Copilot’s Agent Mode and Workspace are noted for their tight integration with GitHub’s pull request and issue-tracking workflow. The tool’s safety and scale positioning appeals to enterprises, and its support for JetBrains, Visual Studio, and Vim makes it the only option for teams standardized on those IDEs.
Slash Commands and Advanced Workflows
The GitHub Copilot app now includes a range of slash commands to streamline workflows. The /plan command helps break down tasks before writing code, while /spar challenges assumptions and identifies potential risks. /autopilot hands off implementation tasks to Copilot, and /rubber-duck uses a different model to independently review work. The /create-canvas command turns conversations into interactive interfaces, and /orchestrate coordinates work across multiple repositories.
These commands are designed to help developers move faster and stay focused. Instead of digging through options, a single / opens up a list of shortcuts. For example, /plan can be used to plan a new feature, prepare for a large refactor, or triage bugs. /spar is useful for validating architecture choices, comparing implementation options, and reviewing migration plans. /autopilot can implement new features or complete larger maintenance tasks.
The /rubber-duck command is particularly valuable for complex work. It uses a different model to independently review plans, refactors, and migration strategies, helping surface blind spots and catch issues the primary model may have missed. This feature is especially useful before opening a pull request or deploying a complex change.
GitHub Copilot’s evolution in 2026 reflects a maturing market for AI coding tools. With over 20 million users, a flexible pricing structure, and a growing list of features, it remains the default choice for many developers. The addition of Kimi K3 and other models provides more options, while internal changes at Microsoft highlight the importance of managing AI usage effectively. As competitors like Cursor and Devin Desktop continue to innovate, Copilot’s distribution advantage and competitive pricing will likely keep it at the forefront of the industry.
