Claude Code Overtakes GitHub Copilot in AI Coding Market Share

The leaderboard flips

For most of the AI coding boom, GitHub Copilot was the default answer to the question of which assistant developers use. That is no longer clearly true. First Page Sage's October 2026 tracking puts Anthropic's Claude Code at 29.4% share as a primary AI coding assistant, up 10.9 points since January. Copilot fell 7.8 points to 22.7% over the same period. 3 The firm says Claude Code passed Copilot in mid-2026 and has widened its lead since. 3

The rest of that table is just as telling. Cursor slipped to 13.1%, down 4.9 points. OpenAI's Codex climbed 7.6 points to 11.8%, nearly tripling its share as developers adopted its cloud task mode. 3 Google's Antigravity and Gemini Code Assist, JetBrains' Junie, and Cognition's Windsurf each sit in the low single digits. 3 The four largest tools now hold 77.0% of the market combined, up from 71.2% in January. 3 That suggests the field is consolidating around a few leaders rather than splintering.

Not everyone counts the same way

The picture depends on what you measure. A developer usage report from Uvik, drawing on JetBrains survey data from January 2026 and Microsoft's FY26 Q1 earnings, still shows Copilot ahead. In that data, Copilot leads at 29% workplace adoption with more than 26 million total users, while Cursor and Claude Code are tied at 18% each. 2 Uvik describes the market as a fractured three-way contest in which different tools win different segments. 2

These figures don't necessarily contradict each other. The Uvik numbers are an early-2026 snapshot of workplace adoption. The First Page Sage numbers track primary tool choice through October. 23 Copilot's huge installed base, helped by its bundling into Microsoft and GitHub workflows, can coexist with developers increasingly reaching for something else as their main tool.

Uvik makes a related point: satisfaction and market share have come apart. Copilot has the biggest footprint, but Claude Code leads in "most loved" sentiment among senior developers. 2 That enthusiasm gap, combined with a terminal-agent interface that suits longer, autonomous tasks, plausibly explains how Claude Code turned goodwill into primary-tool share during 2026.

Uvik also reports that roughly 90% of developers now use at least one AI tool for coding. 2 In a saturated market, growth comes from switching rather than from first-time adoption. That makes Copilot's decline and Claude Code's rise more meaningful than raw user totals suggest.

Codex: the fast follower

Both reports single out OpenAI's Codex as the breakout late entrant. Uvik attributes its traction to how naturally it fits into the ChatGPT ecosystem. 2 First Page Sage credits its cloud task mode. 3 The tool runs across the CLI, an IDE extension, the web, and iOS. On supported ChatGPT plans it now uses GPT-5.5, which OpenAI released on April 23, 2026, with stronger agentic coding and computer-use performance than GPT-5.4. 4 Codex's momentum is the clearest sign that the terminal-and-cloud agent model, not the autocomplete-in-the-IDE model, is where share is moving.

Models are turning over even faster

The models behind these tools change faster than the tools themselves. Developer directories still describe Claude Opus 4.7 as Anthropic's newest top-end model for hard, long-running coding work. 4 Benchmark trackers, however, already list it as superseded. 1

Opus 4.7 scores 87.6% on SWE-bench Verified and 64.3% on SWE-bench Pro, with a 1M-token context window and pricing of $5 in and $25 out per million tokens. 1 That ranks it only fourth on Verified. 1 Anthropic-reported figures show Opus 4.8 reaching 69.2% on SWE-bench Pro. 1 Sonnet 5 posts 85.2% on Verified and 63.2% on Pro, landing within 1.1 points of Opus 4.7 on Pro at a fifth of the input price. 1 Its Verified score also clears Opus 4.6's 80.8%. 1

The Pro scores cited here are run on vendor scaffolds. 1 They are useful for tracking generational progress but are not neutral head-to-head comparisons.

What it means

The takeaway is that the AI coding market is now driven by model quality and agentic workflows more than by distribution. Copilot's incumbency gave it scale, but it did not stop a better-loved, agent-first rival from overtaking it as developers' primary choice. 23

At the same time, the speed of model turnover is a double-edged sword for Anthropic. Cheaper mid-tier models like Sonnet 5 are closing in on last quarter's flagship. 1 Codex is gaining ground on the strength of OpenAI's own model releases. 34 Claude Code's lead is real, but on this evidence it rests on staying a model generation ahead. In a market where a frontier model can be outdated within months, that lead will need constant defending.