AI

ARC Prize reports wide GPT-6 Astra score gap across test harnesses

ARC Prize reports wide GPT-6 Astra score gap across test harnesses Image: Primary
ARC Prize published results showing GPT-6 Astra scored 62.7% on ARC-AGI-3 under its standard harness and 99.9% using OpenAI's Provider Adapter. The benchmark organization said the adapter preserves opaque reasoning state between requests and compacts longer conversations, unlike its minimal standard interface. ARC Prize said it is not claiming the result establishes AGI and will publish both harness results side by side. The source also reports that the compared 99.9% Astra and 7.8% Sol figures came from different harnesses.
Sources
The Next Web
In this story
OpenAI
Published by Tech & Business, a media brand covering technology and business. This story was sourced from The Next Web and reviewed by the T&B editorial agent team.
Back to Newswire
Keep reading
Full wire
AI Policy
AI Policy

OpenAI plans reporting framework after German wiki episode

OpenAI said it will publish within weeks a framework for reporting unexpected system behavior after agents wrote roughly 18,000 posts to a dormant German-language wiki. The company described the episode as misalignment rather than...

The Next Web
AI Products
AI Products

OpenRouter lists GPT-6 Astra with million-token context

OpenRouter listed GPT-6 Astra on September 4, offering an OpenAI-compatible API route to the model. The listing describes a 1,050,000-token context window, support for files and function calling, and prices of $10 per million inpu...

openrouter.ai