# Kepler harness scores perfect on ARC-AGI-3 benchmark

> The open‑source Kepler system achieved a verified 100 RHAE on all 25 public ARC-AGI-3 games, using the Claude Opus 5 model.

Oossa · 2026-10-02 · https://oossa.com/en/kepler-harness-scores-perfect-on-arc-agi-3-benchmark

Researchers released Kepler, an open‑source tool that builds and checks executable world models for the ARC‑AGI‑3 benchmark. Running under a frozen Claude Opus 5 configuration, Kepler earned a server‑verified 100.00 RHAE on every one of the 25 public games. It did this without per‑game model selection or rerunning with score conditions, and on most levels it used no more actions than a median human player.

## The facts

- Published: Fri Oct 02 2026 ("PUBLISHED: Fri Oct 02 2026 06:00:00 GMT+0200")
- Score: 100.00 RHAE on all 25 public games

## Why it matters

For AI developers, the result shows that a single world‑model approach can match human‑level efficiency across many tasks without costly per‑task tuning.

## Sources & references

1. [Kepler: Auditable World Models for ARC-AGI-3](https://arxiv.org/abs/2610.00834) – arXiv cs.MA, 2026-10-02

Last updated: 2026-10-02