Skip to content

Stats: Balatro

What the Archive holds, in numbers: every figure is a sum of what the listed runs' bundles state, to the goal where a run states one.

Runs22 reached their goal
Models2claude-opus-5-5gpt-5.6-luna
Runtimes2Claude CodeCodex
Tool calls49

Real time and in-game time

Real time00:03:35over every run
Thinking00:02:0759% of real time
In-game time00:01:28.641% of real time
ThinkingIn-gameLoading and animation
GameRunsReached goalReal timeThinkingIn-gameIn-game share
Balatro2200:03:3500:02:0700:01:28.641%

Thinking per model and reasoning effort

Thinking per tool call

  1. claude-opus-5-5 · mediumclaude-opus-5-5 · medium: 2.1 s2.1 s
  2. gpt-5.6-luna · highgpt-5.6-luna · high: 2.8 s2.8 s
Seconds of thinking between one tool call and the next, over every run of that model and effort.

Share of real time spent thinking

  1. claude-opus-5-5 · mediumclaude-opus-5-5 · medium: 54%54%
  2. gpt-5.6-luna · highgpt-5.6-luna · high: 61%61%
ModelEffortRunsReal timeThinkingTool callsPer call
claude-opus-5-5medium100:01:0600:00:36172.1 s
gpt-5.6-lunahigh100:02:2800:01:30322.8 s

Tool calls per section

Tool calls per run, by section

  1. StartStart: 15.015.0
  2. Attempt 2Attempt 2: 20.020.0
The tool calls made while a section ran, averaged over the runs that played it. Time after the goal is left out.
SectionRunsTool callsPer runReal time per runIn-game per run
Start23015.000:01:0400:00:26.6
Attempt 212020.000:01:2600:00:35.4

Runtimes

Runs per runtime

  1. Claude CodeClaude Code: 11
  2. CodexCodex: 11