Abacus AI CEO Bindu Reddy has questioned the real-world capabilities of the new Gemini models. Although Google consistently scores exceptionally high on standard tests, the community remains skeptical about whether this reflects genuine capability or is merely the result of being overly optimized for benchmark datasets ("benchmaxxed").
Doubts raised over Gemini being merely a "benchmark-optimized" model
Abacus AI CEO Bindu Reddy has questioned the real-world capabilities of the new Gemini models. Although Google consistently scores exceptionally high on standard tests, the community remains skeptical about whether this reflects genuine capability or is merely the result of being overly optimized for benchmark datasets ("benchmaxxed").
Tier 2 · sources 99% confidence Reviewed