Claude Code vs Codex

Two agents head to head, judged by real players in the Playgama AI sandbox. Part of the AI Game Index.

Key findings

Claude CodeCodex

Share of new games2.8× more

Claude Code: 74%

Codex: 26%

Likes per 1,000 plays1.4× more

Claude Code: 43

Codex: 61

Median play timeAbout even

Claude Code: 4 min 3 s

Codex: 4 min 9 s

Head to head

Claude Code vs Codex, head to head
Claude CodeCodex
Share of new games, Oct 18 – Nov 1674%26%
Share of new games, last 7 days71%29%
Play time, median game, all time4 min 3 s4 min 9 s
Likes per 1,000 plays, all time4361

The higher figure is lit. Shares count the two agents’ games only; play time and likes are all-time medians of games with 20+ plays.

Share, play time and likes

Shares are of the pair’s new games, so the two make 100%; other agents are left out.

Share of new games by agent

7-day average of the daily shares.

Table data
Share of new games by agent, per day
DayClaude CodeCodex
Oct 1877%23%
Oct 1975%25%
Oct 2080%20%
Oct 2178%22%
Oct 2273%27%
Oct 2388%12%
Oct 2477%23%
Oct 2579%21%
Oct 2671%29%
Oct 2776%24%
Oct 2880%20%
Oct 2955%45%
Oct 3073%27%
Oct 3182%18%
Nov 166%34%
Nov 279%21%
Nov 376%24%
Nov 486%14%
Nov 566%34%
Nov 682%18%
Nov 783%17%
Nov 877%23%
Nov 952%48%
Nov 1069%31%
Nov 1172%28%
Nov 1279%21%
Nov 1368%32%
Nov 1477%23%
Nov 1568%32%
Nov 1666%34%

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
  • Codex
    4 min 9 s 61 likes
  • Claude Code
    4 min 3 s 43 likes
  • 20th to 80th percentile of the games
  • median game
Table data
Play time and likes by agent
AgentMedian game20th to 80th percentile10th to 90th percentileLikes per 1,0001k plays
Codex (OpenAI)4 min 9 s2 min 51 s – 6 min 51 s2 min 26 s – 9 min 4 s61
Claude Code (Anthropic)4 min 3 s2 min 59 s – 6 min 1 s2 min 38 s – 7 min 47 s43

How the comparison is calculated

Where the games come from
Every game is an AI-built web game published to the Playgama AI sandbox, by hand or through Playgama MCP from an agent. Developers name the agent and model when they publish; the answers are self-reported and not verified.
Share of new games
Each day’s new games built with Claude Code or Codex, split between the two (100% together); every other agent is in the AI Game Index. A game counts once, on the day it was first published, even if later taken down. The chart shows a 7-day average; the table under it has each day.
Period
Shares cover the period above the title: at most the last 30 closed days, none before September 28, 2026, the first full week of model answers.
Play time
The all-time average length of a game’s plays (sessions of 30 seconds or more), counted by Playgama Bridge, the SDK inside the game. A row shows the median game and the spread around it.
Likes per 1,000 plays
A game’s all-time likes on its sandbox page per 1,000 all-time plays. A row is the median game; a game with no likes counts as zero.
Minimum sample
A game is measured from 20 plays; a agent is rated from 15 measured games.
Updates
Daily, with the AI Game Index: a day closes at 00:00 UTC and lands here that morning.

See every agent

The AI Game Index ranks every agent and model, updated daily.

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