Episode: Hunyuan Hy4 Preview: Open-Source Top Tier
Duration: approximately 7 minutes
Level: B1 (Intermediate)
---
[Mike]: Welcome back to Learn English with Podcasts! Sarah, imagine an AI that could do almost your whole workday - write code, check your invoices, even build a small video game - and this week you can try it for free.
zh:欢迎回到 Learn English with Podcasts!Sarah,想象一个 AI 几乎能搞定你一整天的工作——写代码、核对发票、甚至做个小游戏——而且这周还能免费试用。
[Sarah]: Free? That sounds too good. Which AI are we talking about?
zh:免费?听着太好了吧。我们说的是哪个 AI?
[Mike]: Tencent just released and open-sourced a new model called Hunyuan Hy4 preview. And they say it is now among the very best open-source models in the world.
zh:腾讯刚发布并开源了一个新模型,叫混元 Hy4 preview。而且他们说,它现在已经是全球最好的开源模型之一。
[Sarah]: Wait, Tencent? The company behind WeChat? They made a top AI too?
zh:等等,腾讯?微信那家公司?他们也做出了顶尖 AI?
[Mike]: Yes. And here is the first big number: the model has 770 billion total parameters, but only 49 billion are active each time you use it.
zh:是的。第一个大数字是:模型总参数 7700 亿,但你每次使用时只有 490 亿被激活。
[Sarah]: Only 49 of 770? So it is like a huge company where most people are on standby, and a small team does the actual job?
zh:770 亿里只有 49 亿?那它像一家巨型企业,大部分人待命,只有一小队人真正干活?
[Mike]: Perfect analogy. That design is called Mixture of Experts. You keep a giant library of knowledge, but only wake up the parts you need. It saves cost and speed.
zh:这个比喻太贴切了。这种设计叫专家混合。你保留一个巨大的知识库,但只唤醒需要的那部分。省成本又提速。
[Sarah]: Smart. And what else is impressive?
zh:聪明。还有什么厉害的?
[Mike]: The memory. It can read up to 1 million tokens in one go. That is roughly a whole book, or your entire year of chat logs.
zh:它的记忆力。一次能读多达 100 万 Token。差不多是一整本书,或者你一整年的聊天记录。
[Sarah]: A whole book at once? My notebook would be full after two pages.
zh:一次读完一整本书?我的笔记本两页就满了。
[Mike]: And because it is open-source, developers can download the weights, study how it works, and run it on their own servers.
zh:而且因为它开源,开发者可以下载权重、研究它的原理,还能部署在自己的服务器上。
[Sarah]: So it is not locked inside one app. The whole world can build on it?
zh:所以它不被锁在某个 App 里。全世界都能在它基础上再创造?
[Mike]: That is what open-source means. Tencent shared it on Hugging Face, GitHub, and ModelScope.
zh:这就是开源的含义。腾讯把它放到了 Hugging Face、GitHub 和 ModelScope 上。
[Sarah]: Okay, but is it actually good, or just open?
zh:好,但它真有本事,还是只是开源而已?
[Mike]: Tencent built it for work, not small talk. They tested it in a blind test with 163 internal experts across 203 engineering tasks.
zh:腾讯是冲着干活造它的,不是闲聊。他们做了盲测,请了 163 位内部专家,覆盖 203 个工程任务。
[Sarah]: Blind test meaning the experts did not know which model they were grading?
zh:盲测就是说专家不知道自己在评哪个模型?
[Mike]: Exactly. And Hy4 preview scored 2.99 out of 4, just above GLM 5.3 at 2.92 and Kimi K3 at 2.94.
zh:没错。Hy4 preview 拿到 4 分里的 2.99,略高于 GLM 5.3 的 2.92 和 Kimi K3 的 2.94。
[Sarah]: So it barely edged them out. Close race at the top.
zh:所以它只是险胜。顶尖这场很胶着。
[Mike]: Very close. Now for real examples. In office work, it read 72 invoices and three company rule books, then decided which reimbursements were allowed.
zh:非常接近。说点实际例子。办公场景里,它读了 72 张发票和三本公司规章,然后判断哪些报销能过。
[Sarah]: Three rule books? That alone would take me an afternoon.
zh:三本规章?光这个我就得耗一个下午。
[Mike]: And in game development, it connected to the Unreal 5 engine through something called MCP, and built a playable shooting game just by talking to it.
zh:游戏开发里,它通过一个叫 MCP 的东西接入了 Unreal 5 引擎,光靠对话就做出了一款能玩射击游戏。
[Sarah]: A whole game from conversation? No coding by hand?
zh:整款游戏靠对话生成?不用手写代码?
[Mike]: That is the claim. The developer just kept chatting to improve it. And in science, it ran a molecular simulation of 32,512 atoms at 54.9 milliseconds per step.
zh:这是它的说法。开发者只要不断对话来完善。科学方面,它跑了一个 32,512 个原子的分子模拟,每步 54.9 毫秒。
[Sarah]: Each step takes just a few milliseconds? That sounds fast for something so small.
zh:每步只要几毫秒?那么小的体系还这么快?
[Mike]: They say it is 2.0 times faster than before, and one top GPU could hold 300,000 atoms. Useful for drug and material research.
zh:他们说比以前快 2.0 倍,一张高端 GPU 能装下 30 万个原子。对药物和材料研究很有用。
[Sarah]: Okay, but here is the part that gave me chills. This model helped improve itself?
zh:好,但最让我起鸡皮疙瘩的是这点。这个模型还帮着自己改进自己?
[Mike]: Yes. Hy4 preview joined its own development - it suggested training methods, data plans, even optimized the computer code that runs it. End-to-end speed went up 31.8 percent.
zh:是的。Hy4 preview 参与了自己的研发——它提议训练方法、数据方案,甚至优化了运行它的底层代码。端到端速度提升了 31.8%。
[Sarah]: A model that tunes its own engine. That is like a car that redesigns its own engine while driving.
zh:一个会调自己引擎的模型。就像一辆车一边开一边重新设计自己的引擎。
[Mike]: And Tencent has been rebuilding its AI infrastructure since February, shipping a big update about every two months. They release a preview first, then a final version.
zh:而且腾讯从 2 月起重建了 AI 基础设施,大约每两个月发一个大版本。他们先发预览版,再发正式版。
[Sarah]: So the preview is like a beta you get to use early?
zh:所以预览版就像你提前用上的测试版?
[Mike]: Exactly. And to collect feedback, WorkBuddy and CodeBuddy are free to try for two weeks. If you want, you can build something with it today.
zh:没错。为了收集反馈,WorkBuddy 和 CodeBuddy 限时两周免费。想试的话,今天就能用它做个东西。
[Sarah]: I like that. Listeners, if you tried Hy4 preview, what would you build first - a game, a report, or your own little assistant?
zh:这个我喜欢。听众们,如果你试了 Hy4 preview,会先做什么——游戏、报告,还是你自己的小助手?
[Mike]: Thanks for listening to Learn English with Podcasts. The next time you hear a model is big, ask not how big, but how much of it actually shows up to work.
zh:感谢收听 Learn English with Podcasts。下次听说某个模型很大时,别问它多大,问问它到底有多少真的来上班了。
[Sarah]: See you next time!
zh:下次见!
