# Simple RL and Graph Search Outperforms GPT-5.4 by 30x on ARC-AGI-3

- slug: simple-rl-and-graph-search-outperforms-gpt-54-by-30x-on-arc-agi-3
- date: 2026-04-06
- category: Artificial Intelligence

A simple RL+graph-search baseline — not a frontier LLM — scored 12.58% on ARC-AGI-3, outperforming every commercial model by 30x. GPT-5.4 hit 0.26%. Gemini 3.1 Pro: 0.37%. Humans: 100%.

---