Skip to content

ARC-AGI-3 Semi-Private: Provider Adapter harness

ARC Prize verified interactive reasoning results using a provider-specific stateful adapter, kept separate from Standard harness results.

8 models

Higher is better.

  1. 1GPT-6 AstraOpenAI98.55
  2. 2Gemini 4 ArgonGoogle97.30
  3. 3GPT-6.1 SolOpenAI96.18
  4. 4Gemini 3.8 FlashGoogle35.00
  5. 5GPT-6 SolOpenAI23.03
  6. 6Grok 4.7xAI7.02
  7. 7Claude Haiku 5.5Anthropic2.82
  8. 8GPT-6 LunaOpenAI0.59
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Authors
ARC Prize Foundation (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.