Vibeleaderboard

Vibers Apps Tools Intel

Index — Latest Intelligence

Intel

© 2026 Vibeleaderboard

Privacy Terms Submit Map API for Agents

Two settings tripled GPT-5.6 Sol’s ARC-AGI-3 score — Intel — VibeLeaderboard

Intel / article

Two settings tripled GPT-5.6 Sol’s ARC-AGI-3 score

Source: OpenAI
Author: OpenAI
Date: Jul 30, 2026

Why it matters

The result is a clean warning against reading benchmark scores as model-only measurements: memory, context management and harness design can dominate the apparent capability.
OpenAI found that retaining reasoning and replacing rolling truncation with compaction moved GPT-5.6 Sol from 13.3% to 38.3% on the ARC-AGI-3 public set while using six times fewer output tokens.

Transcript

OpenAI re-ran ARC-AGI-3 through the Responses API with retained reasoning and compaction. The same GPT-5.6 Sol model moved from 13.3% to 38.3% on the public set while producing six times fewer output tokens.

Read the source ↗openai.com

More from OpenAI

articleGPT-5.6 Sol helped optimize the infrastructure that serves it
postWe’re sharing new research with @apolloaievals on reward-seeking—when models fol

Recommended reads

articleGPT-5.6 Sol helped optimize the infrastructure that serves itOpenAI
article[AINews] Claude Opus 5: Fable-level performance at Opus price (half Fable)www.latent.space
articleGPT-5Ethan Mollick

Comments

Sign in to comment.

Loading comments…