Union Alpha uses 3x fewer tokens than Ox Alpha on the same 3D siege task
On the same Three.js procedural 3D robot siege task, Union Alpha used 44,358 tokens at $0 versus Ox Alpha's 127,064 tokens at $0.57, with the author saying Union writes code directly while Ox thinks first and is more verbose; Ox Alpha has been revealed as GLM-5.3-Flash.
💥 Union Alpha 对比 Ox Alpha 按在地上摩擦 同一套 3D 机器人军团攻城任务(纯 Three.js 程序化生成,无贴图无模型),结果几乎一边倒: 🔹Token 消耗:Union Alpha 44,358 vs Ox Alpha 127,064(少了近 3 倍) 🔹成本:Union Alpha $0 vs Ox Alpha $0.57 🔹风格:Union 直接写代码,Ox 先想再写,所以更贵、更啰嗦 🔹画面:Union 的激光束在画面中途切开大楼,整场最电影感的一镜来自免费模型 🔹定位:Union 为 Agent 编程而生,支持图片,且明确不拿用户数据训练 Ox Alpha 是已经被揭晓的 GLM-5.3-Flash。Union Alpha 还是隐身模型,但这一局已经说明:agentic coding 拼的不是“想得有多长”,而是“代码写得有多准”。 免费窗口只有一周,趁现在去 http://aimlapi.com 试。 #UnionAlpha #OxAlpha #StealthModel #AgenticCoding #OpenCode #AIMLAPI #AI编程 #ThreeJS https://x.com/aimlapi/status/2100252462288339273?s=46
This post quotes another post: x.com/aimlapi/status/2100252462288339273