---
title: "Enabling Two Settings Triples ARC-AGI-3 Benchmark Scores"
source: "OpenAI：News（RSS）"
source_tier: "t1"
url: "https://openai.com/index/how-two-settings-tripled-our-arc-agi-3-scores"
permalink: "https://feed.astralor.com/items/itm_cc8e132c619995d199d33807"
published: "2026-07-29T15:00:00.000Z"
score: 91
channel: "tech"
tags: ["ARC-AGI-3","GPT-5.6","OpenAI","Compression","Reasoning Retention","AGI"]
---

# Enabling Two Settings Triples ARC-AGI-3 Benchmark Scores

## AI commentary

> This finding is crucial! It shows that model performance depends not only on training but also on inference-time configuration. For AI engineers and researchers, this is a must-know optimization direction. Teams deploying models on resource-constrained devices in particular could find a low-cost way to boost capabilities.

## At a glance

OpenAI announced that by enabling two API settings—Reasoning Retention and Compaction—its GPT-5.6 model tripled its scores on the ARC-AGI-3 benchmark while also improving inference efficiency. This finding reveals the significant impact of inference-stage configuration on complex general intelligence tasks, offering a new practical direction for AI system optimization.

## Key points

- Reasoning Retention maintains a more complete logical chain during inference.
- Compaction reduces redundant computation by streamlining output structure.
- Combining both yields a three-fold score increase on ARC-AGI-3 with improved efficiency.
- The result offers a low-cost, high-return strategy for deployment tuning of AI models.

## Full text

How two API settings improved GPT-5.6 performance on ARC-AGI-3, boosting scores and efficiency by retaining reasoning and enabling compaction.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_cc8e132c619995d199d33807