ModelsMulti-source

Qwen4-Max 基准出炉:Artificial Analysis 拿 43 分,原生视觉加 1M 上下文

Why we picked it

阿里新版旗舰 Qwen4-Max 的成绩单来了,比上一代高 9 分,智能体会话实测也有 +7.5% 提升。做 Agent 或长文本应用的开发者可以对照 Kimi K3.1 Pro 的分数看看值不值得切换。

阿里巴巴 Qwen4-Max 的基准成绩公布,在 Artificial Analysis Index 拿到 43 分,比 Muse Wonder 低 1 分,比 Kimi K3.

Body

1 Pro(46 分)低 3 分,但比 Qwen3.8 Max(xhigh)高出 9 分。

模型采用新的 MoE 架构,支持原生视觉输入,上下文窗口达到 1M token。在 xhigh reasoning effort 设置下跑出的成绩显示,Qwen 系列相比上一代提升明显。

Agent Arena 是另一个亮点:Qwen4-Max 首次上榜即排名第 14,在 14.7k 个真实世界智能体会话中取得 +7.

5% 的净提升,说明其在多步任务执行上有实际表现。与 Kimi K3.

1 Pro、Muse Wonder 同场对比可以看出,顶级旗舰档的分数差距已收敛到个位数。

Original · TestingCatalog

ALIBABA 🔥: Qwen4-Max benchmarks are out, and it scored 43 points on the Artificial Analysis Index, one point below Muse Wonder.

Qwen4-Max is a new MoE architecture model with native vision and a 1M-token context window.

> Qwen4 Max, using xhigh reasoning effort, scores 9 points ahead of Qwen3.8 Max (xhigh) and 3 points behind Kimi K3.1 Pro (max, 46).

> It also debuted at #14 overall on Agent Arena with a +7.5% net improvement across 14.7k real-world agentic sessions.

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Models" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison and AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Qwen4-Max, Alibaba, Kimi K3.1 Pro, MoE.

Source attribution: curated from TestingCatalog (@testingcatalog). Cross-source corroboration: 9 independent sources have covered this development (including 阿里通义 Qwen, lmarena.ai, OpenRouter), confirming industry-wide signal strength.