DeepSeek V4.1 Flash 1017 Is WILD… 91.5% in 251 Tests for Just $1.80!
DeepSeek just launched V4.1 Flash 1017, an update pitched as better scores, faster task completion and shorter outputs, and it’s open weights (MIT, ~763B, 1M context). I ran it through my own 251-test suite on launch day, twice: the first run scored 84.2%, and a rerun with a bigger token budget lifted it to 91.5%.
影片由 AI WITH Rithesh 發布。觀看原始影片。
原始發布日期：。片長：11 分 39 秒。
相關文章：DeepSeek V4.1-Flash-1017 是什麼？10 月 17 日改版內容、官方跑分與 API 價格一次看懂。