Gemini 4 Carbon 擊敗 GPT-6.1 Sol 奪 MathArena 榜首，但成本高 15 倍
谷歌的 gemini 4 carbon 在 matharena 數學基準測試中擊敗 gpt-6.1 sol，但成本高 15 倍。這影響了優先考慮數學任務性能的用戶。
google's gemini 4 carbon model has become the first non-openai model to top the matharena math benchmark, achieving an 85.4% score with a cost of $15.70. it surpassed openai's gpt-6.1 sol, which scored 80.6% at a cost of $1.07. anthropic's claude fable 5.2 came in second, scoring 82.8% at a cost of $40.20. gemini 4 carbon's high performance comes with a higher cost, 15 times that of gpt-6.1 sol. this development affects users prioritizing math task performance and may influence their model choices.