Google's Gemini 4 Argon scores 97.5% on ARC-AGI-3 under a vendor-specific harness
Original titleGoogle's Gemini 4 Argon is closer to the AI frontier than any Google model in months
AISummary
Google DeepMind's Gemini 4 Argon scored 97.5 percent on ARC-AGI-3 at a cost of $4,800 using a vendor-specific Provider Adapter Harness, according to ARC Prize. Under the standard harness, Argon scored 49.8 percent at low reasoning, and Epoch AI places it near the frontier on its Capabilities Index but behind on math and software engineering.
AIWhy it matters
The article explains how the benchmark score depends on the testing harness, which matters for judging the reported ARC-AGI-3 result.
Source: The Decoder · the-decoder.comPublished · added here