Google's Gemini 4 Argon Nears AI Frontier with Record Benchmark Score
Google DeepMind's Gemini 4 Argon achieved a notable 97.5 percent on the ARC-AGI-3 benchmark under specific conditions, positioning Google alongside OpenAI past the 95 percent threshold. However, performance varies significantly based on test setups, with standard harness scores dropping to 49.8 percent, while the model still lags competitors in mathematics and software engineering capabilities.
Source: The Decoder