Google’s newly released Gemini 3 Flash AI model has shown outstanding performance in tests, completely surpassing the previous 2.5 Pro version. In the 100K attention test, it achieved a 100% recall rate, and its visual testing performance is on par with the 3 Pro model. Development and tuning data show significant improvements in both reasoning speed and accuracy, especially demonstrating high efficiency in the needle-in-a-haystack test. Based on manual input testing through official channels, the model excels in multiple scenarios, bringing a new breakthrough to the AI field and signaling continuous advancement in AI technology.
Original Link:Linux.do
评论前必须登录！
立即登录 注册