Playgama AI Game Index: 8,719 games show Claude Code leads volume, Codex wins more likes
By GameSlash
November 12 Playgama survey data covers 8,719 agent-built web games and 2.17M launches, revealing different game shares, play times and likes.
A November 12, 2026 snapshot of Playgama's AI Game Index gives one of the clearest public comparisons yet of games published through AI coding agents: 8,719 games and 2.17 million launches. Claude Code was credited for 49% of those games and Codex for 18%, while Codex-associated games received about 1.4 times as many likes per 1,000 plays in that snapshot. The index itself first launched in late October; the news is the larger November data set and the comparison published on November 12, not a newly launched platform.

Agent popularity does not equal game quality
In the November 12 snapshot, Claude Code accounted for 49% of submitted games, Codex for 18%, and the Claude desktop/web app for 12%. The reported rate of likes was 57 per 1,000 plays for Codex games, compared with 42 for Claude Code games. This association does not prove Codex creates higher-quality games. Creators identify their own AI agent and model on submission, and Playgama does not verify those declarations.
Session length was close for large agents: about 4 minutes 8 seconds for Codex-associated games and 4 minutes for Claude Code games. Gemini CLI had the longest median reported game session at 4 minutes 29 seconds, but the sample was small: about 0.5% of submitted games. An agent or model enters the ranking only after at least 15 games with at least 20 player launches each.
The weekly source shows how agent share shifts
Playgama also keeps a separate week-by-week view. For November 2–8, the weekly index reports Claude Code on 49% of new games and Claude Opus as the model on 68%. Those are weekly shares of newly submitted games, and should not be confused with the cumulative percentages from the November 12 snapshot.

How to use this responsibly when building with AI
The dataset covers games built and published through Playgama's AI sandbox, not games across Steam, itch.io, or the industry as a whole. Its like and play-time metrics also depend on players, genre, distribution, and sample sizes. A creator comparing coding agents should use these observations as hypotheses, then test the same gameplay prototype with different agents and collect actual completion, playtime, and user feedback rather than reading the leaderboard as an independent benchmark.
See Playgama's live AI Game Index and methodology and the November 12 GameDev Reports analysis. GameSlash has previously reported on Playgama Sandbox and publishing with AI agents.