OpenAI's new GPT-5.6 model, specifically the Sol version, has achieved a score of 7.8% on the ARC-AGI-3 benchmark, a task designed to challenge AI's fluid intelligence. While this score is significantly low compared to human performance (over 90%), it represents a massive leap from previous AI models, with GPT-5.5 scoring only 0.43% and Anthropic's Opus 4.8 scoring 1.5%. The model also demonstrates high efficiency, achieving over 90% on earlier ARC-AGI versions. AI
IMPACT Demonstrates significant progress in fluid intelligence benchmarks, potentially accelerating AI capabilities in complex reasoning tasks.
RANK_REASON Frontier-lab model release with system card and benchmark results. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
Read on The Algorithmic Bridge (Alberto Romero) →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
