Qwen 发布 Qwen4-Max:原生视觉与 1M token 上下文
Qwen 又出新旗舰 Qwen4-Max,1M 上下文加上原生视觉,长文档和图片任务一把抓。基准上赢了 Opus 5.5 和 GPT-6.1 Sol/Astra,不过还打不过 Fable 5.2。做长文本应用的朋友可以先拿自己数据实测一下。
阿里巴巴 Qwen 团队推出新模型 Qwen4-Max,具备原生视觉能力,上下文窗口达到 1 百万 token。据发布者称,该模型在相关基准测试上超过 Opus 5.5 和 GPT-6.1 Sol/Astra,但仍落后于 Fable 5.2。原生视觉意味着模型可直接处理图像输入,而非依赖外挂视觉模块。发布者也提醒,基准成绩与真实使用体验之间存在差距。
对开发者而言,1M token 上下文适合长文档分析、大规模代码库理解和多文件重构等场景;原生视觉则减少在图像理解任务中拼接多模型的工程成本。目前该模型的具体基准名称、定价和可用渠道尚未在原文中披露,实际表现还需结合自身任务验证。
Interesting release: Qwen has launched Qwen4-Max, their new AI model with native vision and a 1 million-token context window.
Outperforms Opus 5.5 and GPT-6.1 Sol/Astra on relevant benchmarks bot lacks behind Fable 5.2. But as we know, numbers and real world use are two different things.
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 Qwen, Qwen4-Max, long context, multimodal.
Source attribution: curated from kimmonismus (@kimmonismus).