AI 导读
在我们的语音推理基准 Big Bench Audio 上，Gemini 3.8 Live Extended Thinking (High) 得分 97.7%，领先于 Grok Voice Think Fast 2.0 High（97.2%），落后于 Qwen Audio 3.0 Realtime Plus 的 99.2%。标准版 Gemini 3.8 Live 得分为 91.7%。https://t.co/aatRk9efWw
正文
On Big Bench Audio, our speech reasoning benchmark, Gemini 3.8 Live Extended Thinking (High) scores 97.7%, ahead of Grok Voice Think Fast 2.0 High (97.2%) and behind Qwen Audio 3.0 Realtime Plus at 99.2%. The standard Gemini 3.8 Live scores 91.7%. https://t.co/aatRk9efWw
来源：@ArtificialAnlys · x.com