{
  "id": 6620124,
  "title": "Astra's Adapter Beat the Reasoning Dial on ARC-AGI-3",
  "url": "https://urgent.news/2026/09/10/astras-adapter-beat-the-reasoning-dial-on-arc-agi-3",
  "topic": "culture",
  "section": "Culture",
  "published": "2026-09-10T18:20:09.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/reidmarlow/astras-adapter-beat-the-reasoning-dial-on-arc-agi-3-272e"
  },
  "original_language": "en",
  "account": "On 3 September 2026, ARC Prize released the results for Astra's ARC-AGI-3 model. In the Standard harness with maximum reasoning capabilities, the model achieved a score of 62.7% and cost $26,098 to run. When the Provider Adapter was used with high reasoning settings, the model scored an impressive 99.9% but incurred a lower cost of $18,817 for the run. Though the scores were derived from separate harness configurations, both results were confirmed as genuine. The Standard harness allows the model to retain chosen notes for future reference, while the Provider Adapter maintains an opaque reasoning state across requests, enabling it to reuse previously executed work. Despite sharing the same underlying weights, the differing memory management approaches resulted in distinct scores.",
  "summary": "On 3 September 2026, ARC Prize published GPT-6 Astra's ARC-AGI-3 numbers. Greg Kamradt's post puts two scores on the same model. In the Standard harness at max reasoning, Astra hit 62.7% and the run cost $26,098. In OpenAI's Provider Adapter at high reasoning, it hit 99.9% and the run cost $18,817. I am quoting the ARC Prize table , not a launch slide. The official results page lists twelve…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}