Event ·Multi-source

腾讯 Hy4 发布:Artificial Analysis 智能指数得 35 分,为中国实验室第二名模型

腾讯混元发布 Hy4,是今年早些时候 Hy3 的后继版本,在 Artificial Analysis Intelligence Index v5 得 35 分,与 Grok 4.7 (xhigh) 持平,比 Claude Haiku 5.5 (max) 低 1 分。在中国实验室模型中,Hy4 排名第二,仅次于得 46 分的 Kimi K3.1 Pro。智能体编码方面,Hy4 在 FrontierCode 2.0 得 31%,超过 DeepSeek V4.1 Flash 1017 (max) 的 28% 和 Kimi K3.1 Pro 的 25%。定价为每百万输入/输出 tokens $0.834/$2.501,单位智能成本比 Grok 4.7 便宜 10 倍。短板是幻觉率高达 89%,而 Kimi K3.1 Pro 为 37%。

Current read

腾讯混元出了新模型 Hy4,跑分和 Grok 4.7 打平,编码能力超过 DeepSeek 和 Kimi,但幻觉率 89% 有点高,价格倒是便宜。

10 sources72° AI heatLast updated 2026/11/5 08:28:08

Evidence

10 sources
Primary sourceArtificial Analysis
View source
Related source 2shao__meng
View source
Related source 3kimmonismus
View source
Related source 4lmarena.ai
View source
Related source 5Nous Research
View source
Related source 6CNA: Business
View source
Related source 7Nikkei Asia
View source
Related source 9rohanpaul_ai
View source

Conflict 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.