Gemini 3.8 Flash just dropped, and 305 tokens per second is hard to ignore
热度趋势
趋势数据积累中
百分比基于当前可用热度信号,而非评论数或独立用户人数。
OpenAI 相关模型动态已经出现,适合跟踪能力变化、生态影响和后续可用性。
Gemini 3.8 Flash 已发布,其输出速度高达每秒 305 个 token,这一速度令人瞩目。这几乎是排名第二的 Muse Spark 1.2(每秒 154 个 token)的两倍,也远超 GPT-5.6 Luna 的每秒 126 个 token。Gemini 3.8 Flash 的智能得分是 59,与得分在 60 到 66 之间的模型组接近。如果这种速度在正常的 API 使用中得以保持,那么长时间的编码代理运行将会变得更加流畅和顺畅。
The speed chart is what got me. Gemini 3.8 Flash is listed at 305 output tokens per second, almost twice the 154 shown for second-place Muse Spark 1.2 and well ahead of GPT-5.6 Luna at 126. Its intelligence score is 59, close to the group sitting between 60 and 66. If that speed holds up in normal API use, long coding-agent runs could feel a lot less painful.
Source: https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/