On 28 August Tencent officially released and open-sourced Hunyuan Hy4 preview. Total parameters 770B, context beyond 1M, input 6 yuan per million tokens — cheaper than DeepSeek. What does this large model price war mean for ordinary people and content creators?
First, understand the news: the one-sentence version
Tencent has open-sourced Hy4 preview, its newest large model, positioning it around productivity — that is, writing code, making spreadsheets, analysing data and doing research. The key figures:
- Total parameters 770B, activated parameters 49B — enormous in size, but only a small part is activated at runtime, so it is fast and cheap
- Context 1M — equivalent to reading the whole Three-Body trilogy in one go
- Input 6 yuan per million tokens, output 18 yuan per million tokens — cheaper than the current mainstream models
- Blind tests beat peers: internal blind testing with 163 experts and 203 engineering tasks gave an average of 2.99, slightly ahead of GLM-5.3 (2.92) and Kimi K3 (2.94)
- Open source and launched at the same time: Tencent Cloud TokenHub, OpenRouter, and products including WorkBuddy, CodeBuddy, Yuanbao and ima
In one sentence: a stronger model, a lower price, and open sourced so you can use it freely.
Why is this worth your attention?
Many people think another large model release has nothing to do with them. But this time is different; it hits three points ordinary people care about most:
1. The price war has reached deep water, and using AI keeps getting cheaper
Last year, running a task on a top model cost several tenths of a yuan each time. Now input has fallen to 6 yuan per million tokens — what does that mean?
Let me work it out: a 2000-word article uses roughly 3000-4000 tokens. Running it once on Hy4 costs less than 3 fen. Writing one article costs you less than a tenth of a yuan.
In the past it was that AI is powerful but I cannot afford it; now it is that AI is so powerful I do not know how much money to make with it.
2. Open source means there is no ceiling
Open source means you can deploy it yourself, tune it yourself and are not restricted by the vendor. Companies can deploy it privately to protect data; individual developers can build on top of it.
Remember the excitement when DeepSeek open-sourced? Hy4 being open source this time adds another log to the industry fire.
3. Tencent pushed AI into products you use every day
Hy4 is not a castle in the air; it is already in WorkBuddy, CodeBuddy, Yuanbao and ima. These are Tencent's own products, which means Tencent has polished the AI capability for you and you simply use it.
For ordinary people, where is the real opportunity?
I am not selling anxiety; here are three solid directions:
Direction one: content creators, treat AI as cheap productive power
When you use AI to write, make images and edit video now, what hurts most? The token cost.
Hy4 presses the cost to 3 fen an article, so you can confidently let AI help you:
- produce topics in bulk, write first drafts, rewrite headlines
- analyse hot topics and interpret data
- rewrite content across platforms (official account to Zhihu to Weibo)
It used to be that AI is too expensive to use often; now it is that AI is so cheap you might as well use it.
Direction two: office workers, outsource the grunt work to AI
Writing reports, making slides, tidying data, cross-file analysis — the work that eats your time is exactly what Hy4's office analysis capability matches. Tencent specifically optimised data analysis and cross-file collaboration, carrying you from information processing to document delivery.
The time you save is your capital for a side business or a new skill.
Direction three: founders and developers, seize the open-source dividend
The greatest value of an open-source model is that it can be deployed privately. If you have a data-sensitive business idea (medical, financial, legal), Hy4 being open source lets you build an AI foundation at low cost.
Especially in the medical industry — data privacy is a red line, and a privately deployed AI assistant is a clear industry need.
My judgement (from the view of a content founder who practises AI)
From the perspective of someone who uses AI, the winner of this round of model competition is not any single company, but every person who can afford to use AI.
Three years ago top AI was a luxury; today it is a daily necessity. And the endgame of the price war is certainly that AI capability becomes as common as water and electricity — by then the gap between people who can use AI and people who cannot will be at its widest.
The smartest thing to do now is one thing: start using Hy4 immediately and get fluent with this AI lever at the lowest cost. By the day it becomes universal, you are already ahead.
#AI #large models #Tencent Hunyuan #side income #making money