Tencent Hy4 Flash: A $0.07 AI Model That Does Everyday Work for a Fraction of the Cost

Tencent announced on November 16, 2026, a new AI model called Hy4 flash — and its selling point is not raw power. It is price. Aimed squarely at everyday work such as research, slide creation and spreadsheet editing, Hy4 flash costs a fraction of Tencent’s flagship model per task, while keeping the million-token context window and tool-using skills that make it useful for real jobs.

The launch comes less than two weeks after Tencent’s original Hy4 model arrived on November 4, and it lands in the middle of a fast-moving week for AI models. Alibaba’s Qwen4-Max launched the same weekend as a new flagship, OpenAI’s GPT-5.2 is rolling out to professional users, and NVIDIA just unveiled Nemotron for Science for research agents. Tencent’s answer is different: instead of chasing the top of the benchmark charts, it is chasing the bottom of the price list.

For anyone who uses AI daily — drafting documents, summarizing research, building slides, cleaning up spreadsheets — that trade-off is worth a closer look.

What Hy4 Flash Actually Is

Hy4 flash is the cheaper, faster tier of Tencent’s Hy4 model family. It has 224 billion total parameters, including 64 billion Engram parameters, with only 12 billion active parameters per forward pass. That sparse design is what keeps the cost down: most of the model’s capacity stays idle on any given request, and only a small slice does the work.

It supports multimodal inputs, so it can take in images as well as text, and it has a context window of up to one million tokens — the same capacity as the full Hy4 model. In practical terms, that means you can feed it a long report, a stack of research papers or a large spreadsheet and ask it to work across the whole thing without it forgetting the beginning.

Tencent lists image understanding, structured output, function calling, caching and deep reasoning among its capabilities. Those are the ingredients of an everyday work assistant: read a document, call a tool, return a cleanly formatted answer, and remember what it was thinking.

The Price: 95% Cheaper Per Task

Illustration of AI token prices dropping sharply in a descending bar chart

The headline number is the price. Tencent Cloud’s published pricing lists Hy4 flash at $0.07 per million input tokens, $0.281 per million output tokens and $0.003 per million cached input tokens. Compare that with the full Hy4 model, listed at $0.834 per million input tokens, $2.501 per million output tokens and $0.042 per million cached input tokens — roughly twelve times more on input and nine times more on output.

TechNode’s report on the launch puts it even more starkly: on Tencent’s internal WorkIndex benchmark, Hy4 flash cost one-twentieth as much per task as the full Hy4 model — a 95% reduction in cost per completed task.

That benchmark is Tencent’s own, so treat the exact percentage as the company’s claim rather than an independent result. But the list prices are public and verifiable, and they tell the same story on their own: this is a model priced for volume, for the kind of work you run dozens of times a day without thinking about the bill.

There is a launch promotion, too. WorkBuddy users can access Hy4 flash for free through November 30, and Tencent Cloud’s TokenHub is offering API access at half the list price. That makes the next two weeks a cheap time to test whether the model holds up on your own tasks.

What It Can Actually Do

Professional using an AI assistant on a laptop to work on documents, slides and a spreadsheet

A low price means little if the model cannot do the job, so the more interesting question is quality — and here Tencent offers a small piece of outside-style evidence. In a blind test in which 85 WorkBuddy users rated outputs across 259 tasks, Tencent said Hy4 flash scored close to Claude Haiku 5.5 in user satisfaction, while having the lowest cost and the shortest completion time among the models tested.

That comparison matters because Claude Haiku 5.5 is the model most people reach for when they want something fast and cheap. If Hy4 flash genuinely sits in the same satisfaction band at a lower price, it becomes a real alternative for everyday tasks — the document summaries, slide drafts and spreadsheet cleanups that make up most AI usage in offices.

The use cases Tencent names are deliberately ordinary: research, slide creation and spreadsheet editing. Those are not the glamorous frontier tasks that headline model launches usually promise. They are the repetitive, time-consuming jobs that eat up a workday, and a model that does them well at seven cents per million input tokens changes the economics of automating them.

One caveat is worth keeping in mind. The blind test was run by Tencent with its own users, and satisfaction is a soft measure — it captures how people felt about the outputs, not whether the outputs were correct. For work where an error is expensive, the right move is still to test the model on your own documents and check the results before trusting it.

Where You Can Use It

Hy4 flash is available now across Tencent Cloud TokenHub, WorkBuddy, Yuanbao, WeCom, QQ, ima and Marvis, with OpenRouter integration coming soon. That spread covers both the developer route — an API through TokenHub — and the everyday route, through Tencent’s workplace and consumer apps.

Tencent also says it plans to release the model’s weights, which would let developers download and run it themselves rather than paying per token. The original Hy4 model carried the same promise when it launched on November 4, and its weights had not yet appeared by mid-November, so the timing of the flash release is worth watching rather than assuming.

For now, the practical way in is simple: WorkBuddy is free through November 30, and the TokenHub API is half price. If you already use Tencent’s workplace tools, the model is already there. If you do not, the API route is the one to test.

Hy4 Flash vs Hy4: What’s the Trade-Off?

The obvious question is what you give up for a 95% cost cut. Tencent has not published a head-to-head benchmark of the two models on the same tasks, so there is no public scorecard to point to. What the numbers do show is that the two models share the same context window and the same input and output limits — up to 960,000 input tokens and 64,000 output tokens — so capacity is not the difference.

CASRAI’s analysis of the launch makes the key point: matching capacity does not establish matching answer quality, task completion or reliability. Those require evidence from the task being performed, and the only evidence so far is Tencent’s own internal benchmark and its blind user test. The honest way to choose between the two is to run the same real task through both and compare the results — and at Hy4 flash’s prices, that comparison costs almost nothing.

The pattern is familiar from the rest of this month’s launches. The full Hy4 model was positioned for coding, professional work and scientific research, with vision and computer-use skills. Hy4 flash keeps the everyday slice of that — documents, slides, spreadsheets — and drops the price to match. It is the same playbook other labs are running: a flagship for the hardest work, a flash tier for the routine work, and the customer decides which one fits the task.

The Bigger Picture: The AI Price War Hits Everyday Work

Step back, and Hy4 flash is part of a shift that has been building all year. The AI race of 2026 started as a race for capability — bigger models, higher benchmark scores, longer reasoning. It is increasingly a race for cost. Qwen4-Max launched as a closed flagship priced at $1.50 per million input tokens. GPT-5.2 arrived with new pricing tiers for professional work. And now Tencent is offering a capable model at $0.07.

That matters because price decides what gets automated. At flagship prices, AI is a tool you use for important work and think twice about for the rest. At flash prices, it becomes a default — something you run on every document, every spreadsheet, every draft, without doing the mental math first. The companies that win the everyday-work market will be the ones whose models are cheap enough to disappear into the workflow.

There is a risk in that, too. When AI becomes nearly free, it becomes easy to generate far more than anyone checks — more slides, more summaries, more reports, with less human review per item. A cheap model that produces confident-sounding errors at scale is worse than an expensive model used carefully. The cost of the token is not the cost of the mistake.

The sensible approach is the one Tencent’s own launch invites: use the free window and the half-price API to test Hy4 flash on your real work, measure whether the outputs are good enough, and only then let it into the daily routine. At these prices, the test is cheap. The judgment still has to be yours.

Frequently Asked Questions

What is Tencent Hy4 flash?

Hy4 flash is a lower-cost AI model Tencent launched on November 16, 2026, for everyday work such as research, slide creation and spreadsheet editing. It has 224 billion total parameters with 12 billion active per forward pass, supports multimodal inputs, and has a context window of up to one million tokens.

How much does Hy4 flash cost?

Tencent Cloud lists Hy4 flash at $0.07 per million input tokens, $0.281 per million output tokens and $0.003 per million cached input tokens. That is roughly twelve times cheaper than the full Hy4 model on input tokens. On Tencent’s internal WorkIndex benchmark, it cost one-twentieth as much per task as the full Hy4 model.

How does Hy4 flash compare to Claude Haiku 5.5?

In a blind test of 85 WorkBuddy users rating outputs across 259 tasks, Tencent said Hy4 flash scored close to Claude Haiku 5.5 in user satisfaction, with the lowest cost and shortest completion time among the models tested. The test was run by Tencent, so it is the company’s claim rather than an independent result.

Where can I try Hy4 flash?

It is available now through Tencent Cloud TokenHub, WorkBuddy, Yuanbao, WeCom, QQ, ima and Marvis, with OpenRouter integration coming soon. WorkBuddy access is free through November 30, 2026, and TokenHub API access is half the list price during the launch period.

Is Hy4 flash open source?

Not yet. Tencent says it plans to release the model’s weights, but no date has been given. The original Hy4 model launched on November 4 with the same promise, and its weights had not been published by mid-November.

What is the difference between Hy4 flash and Hy4?

Both models share a one-million-token context window and the same input and output limits. Hy4 flash is the cheaper, faster tier aimed at everyday work, while the full Hy4 model is positioned for coding, professional work and scientific research. Tencent has not published a head-to-head benchmark, so the quality gap between them is not yet publicly measured.

The Bottom Line

Tencent’s Hy4 flash is not the most powerful AI model launched this month, and it does not pretend to be. It is the cheapest serious one — a model that handles research, slides and spreadsheets at $0.07 per million input tokens, with a million-token context and tool-using skills, and a blind user test suggesting it holds up against the fast-and-cheap incumbent.

For anyone who uses AI for everyday office work, the launch is worth a test, especially while WorkBuddy is free and the API is half price. The price is the news. Whether the quality holds up on your own documents is the question only a trial can answer — and at these prices, the trial is nearly free.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top