Primary source掘金本周最热
View sourceEvent ·Multi-source
民间AI排行榜单新鲜出炉,Fable 5.1仅排第三
最新民间AI能力排行显示,GPT-6 Astra 瑶瑶领先,GML-5.3 紧随其后,而 Claude Fable 5.1 仅排第三。榜单通过 UP 主设计的贴近现实、条件更苛刻的场景对大模型进行测试,与传统的 MMLU、LiveBench 等学术基准不同,这类测试更接近日常使用场景。传统基准测试的优势在于条件可控、结果可复现,但存在局限,比如无法完全反映模型在复杂、无标准答案的真实任务中的表现。民间实测更接近工作生活场景,但存在样本量有限、主观性较强等问题。Benchmark 跑分领先代表模型在标准化测试场景下实力更强,而民间实测第一代表模型在部分贴近生活、工作的场景中表现更好,二者互为补充,不能互相替代。
Current read
B站UP主做的这个民间测试很有意思,用各种贴近现实、条件更苛刻的场景测试大模型,看它们能不能真正干活,而不是只看跑分有多好看。
11 sources58° AI heatLast updated 2026/9/17 04:51:39
Evidence
11 sourcesRelated source 1DeepLearning.AI
View sourceRelated source 2lmarena.ai
View sourceRelated source 3IT之家
View sourceRelated source 4@OpenAIDevs
View sourceRelated source 5AI Will
View sourceRelated source 6Gary Marcus
View sourceRelated source 7Julien Chaumond
View sourceRelated source 8Justine Moore
View sourceRelated source 9歸藏(guizang.ai)
View sourceRelated source 10Decoder
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.