Models

Ling Flash 3.1 在乐高 Minecraft 任务中击败 Deepseek

Why we picked it

Ling Flash 3.1 在乐高 Minecraft 任务中击败了 Deepseek,展示了其强大的智能体能力。

Ling Flash 3.1 是一个拥有 560B 参数的 MoE 模型,在组装 7,613 个乐高零件的任务中首次尝试就成功完成。

Body

Deepseek 在同一任务中表现不佳。Ling Flash 3.

1 升级了智能体能力,即将发布开源权重。

Original · @atomic_chat_hq

Ling Flash 3.1 beats Deepseek at Lego Minecraft ⛏️

@AntLingAGI's new 560B params MoE model with upgraded agentic capabilities, assembled all 7,613 LEGO pieces first try, while Deepseek struggled with it. Open weights drop soon

Run AI models locally -> https://t.co/RbcCOIgVkj https://t.co/Y0HD4FZpRw

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 AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Ling Flash, Deepseek, MoE, agents.

Source attribution: curated from @atomic_chat_hq (@atomic_chat_hq).