Primary sourceArtificial Analysis
View sourceEvent ·Multi-source
MiniMax-M3.1-Flash 发布: Intelligence Index 得 42 分,单任务成本降至 $0.22
MiniMax-M3.1-Flash 在 Artificial Analysis Intelligence Index 得 42 分,比 MiniMax-M3 高 13 分,与 GLM-5.3-Flash 持平,低于 DeepSeek V4.1 Flash 1017 的 45 分。Humanity's Last Exam 得 48%、CritPt 得 22%,均明显优于 M3 的 39% 和 4%。跑完整套基准只用约 1.89 亿输出 token,单任务成本约 $0.22,不到 M3 的 $0.51 一半。定价为每百万 token 输入 $0.15、输出 $0.60,上下文窗口 1M tokens,支持文本和图像输入。AA-Omniscience 准确率 56%,但因几乎每题都答,幻觉率达 93%。
Current read
MiniMax 新的 Flash 推理模型,跑分追平 GLM-5.3-Flash,单任务成本只要两毛二美元,比上代便宜一半多,10 月 23 号还开源权重。
6 sources55° AI heatLast updated 2026/10/19 15:58:46
Evidence
6 sourcesRelated source 1lmarena.ai
View sourceRelated source 2kimmonismus
View sourceRelated source 3shao__meng
View sourceRelated source 4TestingCatalog
View sourceRelated source 5岚叔
View sourceConflict check
Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.