WHAT NEW IN X10.29 THU

X 上的 AI 合集档案 · 22 个合集 · 58,925 条

deepseek-v4.1VIDEO

DeepSeek V4.1 Flash 与 Kimi K3.1 Pro 同题做 3D 引擎实验室对比

同一提示词在 AgenC Core 里建 Three.js 交互式 V8 引擎实验室:DeepSeek 用 48 分钟、0.86 美元、6720 万 token,Kimi 用 31 分钟、2.16 美元、350 万 token;Kimi UI 更精致,DeepSeek 测功机细节更足。

Sai
@sai0xgen2 天前收录原推发布于 2026.10.25
Sai @sai0xgen原推发布于 2026.10.25
在 X 上打开原推

DeepSeek V4.1 Flash vs Kimi K3.1 Pro tested both models with the same prompt at max effort in AgenC Core: build an interactive 3D V8 engine lab in Three.js. both delivered working simulations with RPM controls, slow motion, exploded views and part inspection. but the resource usage was VERY different > DeepSeek took 48 mins and cost $0.86 > Kimi took 31 mins and cost $2.16 DeepSeek used 67.2M total tokens, while Kimi used just 3.5M. Kimi had the more polished UI and a combustion sequence panel, while DeepSeek built a more detailed dyno cell. one thing to note: DeepSeek was still refining its result when it was stopped, while Kimi finished on its own. which one would you pick? 👀

x.com/sai0xgen/status/2114365812807250224采集于 2026.10.26 12:04