StepFun Step 3.7 Flash 物理竞赛胜 DeepSeek V4-Flash

精选理由

StepFun 的 Step 3.7 Flash 在生成物理动画上把 DeepSeek V4-Flash 比下去了，慢点但模拟和画面都好很多。

AI 摘要

StepFun 的 Step 3.7 Flash 模型与 DeepSeek V4-Flash 在物理动画生成任务中直接对比。任务要求编写自包含 HTML5 Canvas 动画，包含高尔顿板、旋转六边形中弹跳的球、五个同步节拍器三个场景。Step 3.7 Flash 输出 59.6k tokens (9分57秒)，DeepSeek V4-Flash 输出 52.5k tokens (6分21秒)。虽然 DeepSeek 更快，但 StepFun 在物理模拟、视觉效果和逻辑渲染三个维度全面获胜。

原文 · @atomic_chat_hq

StepFun Step 3.7 Flash smashed DeepSeek V4-Flash i…

StepFun Step 3.7 Flash smashed DeepSeek V4-Flash in a physics contest

We gave two open-weight models the same task: write a self-contained HTML5 canvas animation with real physics in one file without libraries. Three scenes - a Galton board, balls bouncing in a spinning hexagon and five metronomes that sync up

Outputs: Step 3.7 Flash: 59.6k tokens, 9m 57s DeepSeek V4-Flash: 52.5k tokens, 6m 21s

DeepSeek was faster, but that's all it had. StepFun's model won on every front: physics simulation, visuals, and logical rendering of each scene

Clement Delangue06-19 00:35原文

查看原推