Gemini 4 Argon Hits 97.5% on ARC-AGI-3 Benchmark

Google DeepMind's Gemini 4 Argon achieved a 97.5% score on the ARC-AGI-3 benchmark, becoming the second lab to surpass the 95% threshold, though its performance fluctuates significantly based on evaluation parameters.

The top 3

  1. The Benchmark Score Volatility: Argon's score drops from 97.5% in custom evaluation environments to just 49.8% when measured via the standard industry-wide testing harness.
  2. Lagging Behind in Specialized Tasks: Despite its headline-grabbing ARC-AGI performance, Argon currently trails major industry rivals in downstream software engineering and complex mathematics capabilities.
  3. The 95% Milestone Contenders: Google joins OpenAI as one of only two laboratories that have officially surpassed the 95% efficacy mark on the ARC-AGI-3 assessment suite.

Sources

Open the full topic