AI Game Index

Which AI agents and models make the games people actually play, judged by real players in the Playgama AI sandbox instead of benchmarks.

Key findings

47%Claude Code of the games in the index came from Anthropic’s agent, more than from any other.

17%Codex of the games came from OpenAI’s agent, the next largest share. Other agents built the rest.

1.4×Likes as many likes per play for Codex’s games as for Claude Code’s: 61 against 43 per 1,000 plays.

1min 11sPlay time separate the rated agents’ longest median game, OpenCode’s 4 min 25 s, from the shortest, ZCode’s 3 min 14 s.

Leaders

TypeAgentModel
Most gamesShare of all games
Claude Code47%then Codex, 17%
Claude Opus58%then Claude Sonnet, 9.1%
Longest playedMedian game
OpenCode4 min 25 sthen WorkBuddy, 4 min 10 s
DeepSeek4 min 14 sthen OpenAI Sol, 4 min 11 s
Most likedLikes per 1,0001k plays
ZCode79then OpenCode, 62
OpenAI Astra64then DeepSeek, 61

AI coding agents compared

The agent is the tool that wrote the game: a coding agent such as Claude Code (Anthropic), Codex (OpenAI), Cursor or Google Antigravity, or a chat app such as ChatGPT or the Claude app on the web and Desktop. The developer names it when they publish.

Share of new games by agent

7-day average of the daily shares. Agents under 2% of all games are counted in All others.

Table data
Share of new games by agent, per day
DayClaude CodeCodexCursorGoogle AntigravityOpenCodeGemini CLIKimi Code AgentGrok BuildWorkBuddyZCodeClaude appChatGPTOther
Oct 1850%15%3.9%3.9%0.5%–0.5%–0.5%–9.7%4.4%12%
Oct 1950%17%2.4%4.4%0.5%0.5%––––11%2.4%11%
Oct 2060%15%1.0%5.4%1.5%0.5%––1.0%0.5%7.8%1.0%6.8%
Oct 2149%14%2.2%4.4%2.2%––0.6%–0.6%13%5.0%8.9%
Oct 2257%21%1.8%0.5%0.5%0.9%––––9.5%1.8%7.2%
Oct 2364%8.7%1.5%1.0%1.0%0.5%––1.9%1.5%14%1.5%4.9%
Oct 2447%14%1.1%6.9%––––2.7%0.5%16%1.6%9.0%
Oct 2554%14%0.5%1.1%1.6%0.5%––3.8%0.5%14%4.3%5.9%
Oct 2652%21%1.5%2.0%0.5%––––0.5%13%3.1%6.6%
Oct 2743%14%4.7%–1.1%0.5%0.5%1.6%1.6%–22%3.7%7.9%
Oct 2853%13%1.0%5.7%2.4%0.5%––0.5%0.5%13%5.3%4.8%
Oct 2937%31%0.5%4.7%–0.5%––3.2%–16%3.2%3.7%
Oct 3047%17%1.7%4.4%1.3%2.2%–0.4%0.9%0.4%17%4.8%2.2%
Oct 3157%12%1.4%5.0%1.4%0.5%––5.0%1.4%10%3.2%3.2%
Nov 145%24%0.5%3.9%0.5%0.5%0.5%0.5%4.8%0.5%7.2%3.4%8.7%
Nov 252%14%0.4%5.8%0.4%0.4%–0.4%3.6%–14%3.6%4.5%
Nov 345%14%–2.7%–1.0%––16%–12%4.0%5.0%
Nov 451%8.4%0.4%1.6%0.4%–4.4%–2.0%0.4%22%2.4%6.4%
Nov 536%19%3.0%0.8%0.4%0.4%3.4%0.4%11%–15%3.0%8.9%
Nov 651%12%1.1%2.8%–1.1%––7.2%–14%3.9%7.2%
Nov 757%11%–1.6%1.1%1.1%–0.5%––15%3.3%8.7%
Nov 848%14%0.5%4.5%1.0%0.5%–0.5%0.5%–20%5.9%4.5%
Nov 931%29%3.8%1.9%2.4%––0.5%–1.4%15%5.7%10%
Nov 1043%19%2.4%2.0%1.2%0.8%–1.6%0.8%–16%3.2%9.7%
Nov 1146%18%0.8%1.2%2.0%–0.4%–3.3%0.8%20%2.5%4.5%
Nov 1245%12%2.8%2.8%1.6%0.4%0.4%0.4%0.4%1.6%16%3.5%13%
Nov 1339%18%1.2%6.4%1.2%0.4%––2.4%0.8%16%2.8%12%
Nov 1442%12%0.4%7.2%2.4%0.4%––0.4%0.4%21%2.8%11%
Nov 1530%14%2.1%11%1.5%0.3%––5.8%0.9%13%7.7%13%
Nov 1641%21%0.7%6.1%1.4%0.7%–0.3%0.3%3.8%16%1.0%8.2%

Play time and likes by agent

Each figure is the agent’s median game among those with 20 plays or more, so one hit cannot carry a row. How it’s measured

Agent Sort by
  • OpenCode
    4 min 25 s 62 likes
  • WorkBuddy
    4 min 10 s 36 likes
  • Codex
    4 min 9 s 61 likes
  • Cursor
    4 min 5 s 43 likes
  • Claude Code
    4 min 3 s 43 likes
  • Claude app
    3 min 58 s 48 likes
  • Google Antigravity
    3 min 52 s 53 likes
  • Gemini CLI
    3 min 30 s 61 likes
  • ChatGPT
    3 min 27 s 51 likes
  • ZCode
    3 min 14 s 79 likes
  • 20th to 80th percentile of the games
  • median game

Not enough data yet to rate Kimi Code Agent and Grok Build. Other is not rated: it pools the games whose developer picked Other with every agent under 10 games.

Table data
Play time and likes by agent
AgentMedian game20th to 80th percentile10th to 90th percentileLikes per 1,0001k plays
OpenCode4 min 25 s3 min 8 s – 8 min 50 s2 min 39 s – 11 min 1 s62
WorkBuddy4 min 10 s3 min 6 s – 6 min 1 s2 min 55 s – 6 min 37 s36
Codex (OpenAI)4 min 9 s2 min 51 s – 6 min 51 s2 min 26 s – 9 min 4 s61
Cursor (Anysphere)4 min 5 s2 min 51 s – 5 min 41 s2 min 39 s – 6 min 41 s43
Claude Code (Anthropic)4 min 3 s2 min 59 s – 6 min 1 s2 min 38 s – 7 min 47 s43
Claude app (Anthropic)3 min 58 s2 min 57 s – 5 min 55 s2 min 38 s – 7 min 14 s48
Google Antigravity3 min 52 s2 min 49 s – 5 min 34 s2 min 27 s – 6 min 33 s53
Gemini CLI (Google)3 min 30 s3 min 12 s – 5 min 51 s2 min 7 s – 6 min 42 s61
ChatGPT (OpenAI)3 min 27 s2 min 43 s – 5 min 2 s2 min 9 s – 7 min 37 s51
ZCode3 min 14 s2 min 41 s – 4 min 17 s2 min 30 s – 5 min 8 s79

AI models compared

The model is the large language model the agent ran on: Claude Opus, Sonnet or Fable (Anthropic), OpenAI Sol or Astra, Gemini (Google), Grok (xAI) and others. Two of them head to head: Claude Opus 5.5 vs GPT-6 Sol and Claude Opus 5.5 vs GPT-6 Astra.

Share of new games by model

7-day average of the daily shares. Models under 2% of all games are counted in All others.

Table data
Share of new games by model, per day
DayClaude FableClaude OpusClaude SonnetClaude HaikuOpenAI AstraOpenAI SolOpenAI LunaOpenAI TerraGemini ProGemini FlashGrokKimiQwenDeepSeekGLMOther
Oct 181.4%70%6.3%–2.1%7.0%–0.7%4.9%3.5%–––1.4%–2.1%
Oct 194.0%59%13%0.7%1.3%7.3%0.7%–6.0%0.7%1.3%––2.0%0.7%3.3%
Oct 204.0%61%9.4%–4.0%7.4%––2.7%4.0%1.3%0.7%–2.0%0.7%2.7%
Oct 216.5%53%12%––12%––2.2%4.3%4.3%––3.6%–2.9%
Oct 2210%57%13%0.6%1.9%9.4%–––1.9%1.9%––1.3%–3.1%
Oct 238.3%68%13%0.7%0.7%2.8%––0.7%1.4%0.7%–0.7%1.4%0.7%1.4%
Oct 245.5%58%6.8%––10%––8.9%0.7%5.5%––2.7%0.7%1.4%
Oct 256.5%63%12%0.7%1.4%7.2%–––1.4%–––3.6%0.7%3.6%
Oct 261.4%66%9.1%–2.8%13%0.7%–0.7%0.7%2.1%–0.7%0.7%–2.8%
Oct 270.7%74%6.9%–0.7%6.9%–0.7%0.7%–3.5%–––0.7%4.9%
Oct 281.8%60%14%1.2%1.8%7.2%–0.6%1.2%4.8%1.2%––3.6%–3.0%
Oct 290.8%55%18%–0.8%13%0.8%0.8%0.8%7.6%0.8%––0.8%––
Oct 302.9%61%12%–7.5%6.9%0.6%–2.9%0.6%0.6%–––0.6%4.6%
Oct 310.6%64%9.4%–1.1%5.6%––2.2%3.3%1.1%–0.6%1.7%0.6%9.4%
Nov 12.2%54%7.9%–1.4%8.6%0.7%3.6%4.3%–2.2%0.7%0.7%5.0%–8.6%
Nov 22.6%61%12%1.3%1.9%6.5%1.3%–3.9%4.5%0.6%––1.9%–1.9%
Nov 30.6%67%13%1.3%5.0%6.9%––0.6%1.9%––0.6%0.6%–3.1%
Nov 40.5%73%8.1%0.5%1.5%1.5%–0.5%–1.5%1.5%5.6%–3.0%1.0%1.5%
Nov 52.2%54%5.0%0.6%1.1%16%–0.6%–0.6%4.5%4.5%–9.5%–1.7%
Nov 6–78%9.9%–1.5%1.5%–1.5%2.3%0.8%––0.8%0.8%–3.1%
Nov 70.8%83%10%––1.7%––0.8%2.5%–––0.8%––
Nov 82.1%65%16%2.1%1.4%3.5%0.7%0.7%2.8%1.4%0.7%––1.4%0.7%1.4%
Nov 9–50%12%–3.9%3.1%17%–1.6%2.3%0.8%––5.4%2.3%0.8%
Nov 105.1%71%8.1%2.2%–5.1%––1.5%2.2%2.2%––0.7%0.7%0.7%
Nov 117.7%67%6.0%–0.5%8.7%–0.5%0.5%0.5%0.5%0.5%–1.1%2.2%4.4%
Nov 127.9%64%6.1%1.2%1.2%10%–0.6%1.8%0.6%1.2%––1.2%3.0%1.2%
Nov 139.7%60%6.3%1.4%1.4%7.6%–0.7%2.1%4.9%0.7%–0.7%2.1%2.1%0.7%
Nov 1410%58%9.5%–1.1%6.1%––5.0%3.9%0.6%––2.2%1.1%2.2%
Nov 153.6%43%6.3%0.9%1.8%9.9%0.5%0.5%5.9%4.5%0.9%0.5%0.5%10%0.9%9.9%
Nov 166.4%51%12%–2.5%9.9%–0.5%2.0%4.4%1.0%––2.5%5.9%2.0%

Play time and likes by model

Each figure is the model’s median game among those with 20 plays or more, so one hit cannot carry a row. How it’s measured

Model Sort by
  • DeepSeek
    4 min 14 s 61 likes
  • OpenAI Sol
    4 min 11 s 55 likes
  • Gemini Pro
    4 min 10 s 43 likes
  • OpenAI Astra
    4 min 9 s 64 likes
  • Claude Opus
    4 min 8 s 43 likes
  • Claude Fable
    3 min 55 s 38 likes
  • GLM
    3 min 41 s 60 likes
  • Grok
    3 min 40 s 36 likes
  • Claude Sonnet
    3 min 38 s 50 likes
  • Gemini Flash
    3 min 27 s 57 likes
  • 20th to 80th percentile of the games
  • median game

Not enough data yet to rate Claude Haiku, OpenAI Luna, OpenAI Terra, Kimi and Qwen. Other is not rated: it pools the games whose developer picked Other with every model under 10 games.

Table data
Play time and likes by model
ModelMedian game20th to 80th percentile10th to 90th percentileLikes per 1,0001k plays
DeepSeek4 min 14 s2 min 49 s – 5 min 59 s2 min 27 s – 11 min 18 s61
OpenAI Sol4 min 11 s2 min 50 s – 7 min 44 s2 min 31 s – 12 min 3 s55
Gemini Pro (Google)4 min 10 s3 min 2 s – 5 min 26 s2 min 40 s – 6 min 21 s43
OpenAI Astra4 min 9 s2 min 58 s – 7 min 11 s2 min 32 s – 8 min 34 s64
Claude Opus (Anthropic)4 min 8 s3 min 1 s – 6 min 11 s2 min 41 s – 8 min 2 s43
Claude Fable (Anthropic)3 min 55 s3 min 1 s – 5 min 47 s2 min 30 s – 7 min 19 s38
GLM (Zhipu AI)3 min 41 s2 min 46 s – 5 min 6 s2 min 40 s – 6 min 32 s60
Grok (xAI)3 min 40 s2 min 47 s – 5 min 48 s2 min 29 s – 7 min 31 s36
Claude Sonnet (Anthropic)3 min 38 s2 min 45 s – 5 min 31 s2 min 31 s – 7 min 6 s50
Gemini Flash (Google)3 min 27 s2 min 41 s – 5 min 4 s2 min 3 s – 6 min 49 s57

How the AI Game Index is calculated

Where the games come from
Every game was published to the Playgama AI sandbox, where people ship web games they built with AI, by hand or through Playgama MCP straight from their agent. When they publish, the developer names the agent and the model they used. The answers are self-reported and not verified.
Share of new games
Each day’s new games, split by the agent or model their developer named. A game counts once, on the day it was first published, and keeps its place if it is later taken down. The chart shows a 7-day average of the daily shares, so a day of a few games cannot swing it; the table under it has each day.
Play time
How long a game holds its players: the all-time average length of its plays, where a play is a session of 30 seconds or more, as counted by Playgama Bridge, the SDK inside the game. A row shows the spread of that figure across its games: the median game and the 20th to 80th percentile, with the 10th and 90th percentiles in its tip and the table. The median, not the mean, so one hit game cannot carry a row.
Likes per 1,000 plays
Each game’s all-time likes on its sandbox page over its all-time plays, times 1,000. A row is the median game, and a game nobody liked counts as zero.
Minimum sample
A game is measured once it has 20 plays. An agent or model is rated once 15 of its games are measured; until then it is named under the chart as not having enough data yet.
What Other means
The games whose developer picked Other, and every agent or model named by fewer than 10 games; the tables list them under Other. Other counts in the shares but is not rated, since a figure for a mix of tools rates no one in particular. On the share chart it joins every agent or model under 2% of all games in one band, All others; the list beside the chart gives each of them its share.
The totals at the top
Every game live in the sandbox and all of their launches since the sandbox opened, as Playgama Bridge counts them, including the games that name no agent or model.
Updates
Daily. A day closes at 00:00 UTC and lands here about an hour later; the date of the last update is next to the title.

Limitations

Frequently asked questions

What is the AI Game Index?

A leaderboard of AI coding agents, such as Claude Code, Codex, Cursor and Google Antigravity, and of the large language models behind them, such as Claude Opus, Claude Sonnet, OpenAI Sol, OpenAI Astra, Gemini and Grok. They are ranked by how real players treat the games they build: how long people play them and how many likes they earn in the Playgama AI sandbox. The ranking is updated daily.

How is it different from coding benchmarks?

Benchmarks such as SWE-bench score code on set tasks, and arenas such as LMArena ask people to vote between two answers. The AI Game Index is a behavioral evaluation: it watches what players do with games that are already published, with no set tasks and no human judges.

Which AI is best for making games?

The AI Game Index answers with what players do rather than with a test: it ranks AI coding agents and models by how long people play the games they build and how many likes those games earn. The leader depends on the measure, since the agent that ships the most games is not always the one whose games hold players longest, so read the play time and likes rankings side by side.

Is Codex better than Claude Code for building games?

Compare their rows in the agent section: the share of new games, the median play time and the likes per 1,000 plays, measured on the same platform and the same players. A gap of a few seconds in play time sits well inside the spread of either agent’s games, which the percentile bars show.

What is the best AI model for vibe coding games?

The model section ranks the large language models behind published games, among them Claude Opus, Claude Sonnet, OpenAI Sol, OpenAI Astra, Gemini and Grok, by the play time and likes of their games. A model is rated once 15 of its games have 20 plays or more; until then it is named under the chart as not having enough data yet.

How is the AI Game Index calculated?

From observed player behaviour, with no human judges. Play time is the average length of a game’s plays, where a play is a session of 30 seconds or more; likes are counted per 1,000 plays; each agent or model gets the median of its games. The methodology lists every threshold.

Where do the agent and model labels come from?

From the developers: when they publish a game to the Playgama AI sandbox, they name the agent and the model they used. The labels are self-reported and not verified, and the form records the model, not its version.

How often is the index updated?

Every day. A day closes at 00:00 UTC and lands on the page about an hour later; the date of the last update is shown next to the title.

How can my game be in the index?

Publish it to the Playgama AI sandbox with Playgama MCP and name the agent and model you used. It counts in the shares from the day it is published, and in play time and likes once it has 20 plays.

What’s next

The index is in beta. More agents, models and measures are on the way as more games ship. Follow Playgama to hear when it grows.

Built a game with AI? Publish it with Playgama MCP and your agent joins the index.