In benchmarks such as MMLU (multidisciplinary questions) and GSM8K (math word problems), it performed comparably to models with significantly higher parameter counts, highlighting xAI’s focus on ...