Make Money with AIHy4freelancingdocument servicesautomationsmall business

Sell a Small Hy4 Document and Computer Task Service at $0.834 Per Million Tokens: Real Costs, Real Prices, No Hype

Tencent launched Hy4 on 4 November, a vision and computer use model at $0.834 per million input tokens. Here is how to price a small document and computer task service on that rate, including the TokenHub discount, without promising income you cannot guarantee.

Published November 5, 2026

Muhammad Usman

By Muhammad Usman · Founder & Lead Reviewer

Sell a Small Hy4 Document and Computer Task Service at $0.834 Per Million Tokens: Real Costs, Real Prices, No Hype

Quick Answer

Hy4, launched by Tencent on 4 November 2026, costs $0.834 per million input tokens and $2.501 per million output tokens, with a calculated $0.417 and $1.2505 on Tencent Cloud TokenHub until 31 December 2026. A small document and computer task service can be priced around those rates, but the discount ends on 31 December, the free WorkBuddy and CodeBuddy quota ends on 15 November, and no income is guaranteed.

What service should you actually sell?

Sell a narrow, repeatable service, not a general offer to automate anything. Hy4's two new capabilities point to three packages that fit well, because each one uses vision, computer use, or both, and each one has a clear delivery the client can check. First, a document reading and cleanup package. A small business gives you a set of scanned documents, charts, or reports, and you hand back clean, searchable text, a summary, and a table of the key numbers, with the source page noted for each figure. Hy4's vision helps here, because it can read charts and document images, but you must still check every number against the source, because vision models misread axes and table rows in quiet ways. The client pays for checked output, not for a raw model dump. Second, a small computer task package. A client gives you a repetitive multi step task inside their own apps, such as moving data between two systems, renaming and filing a batch of files, or filling a standard form from a spreadsheet, and you build and test a supervised workflow that does it, with a written runbook the client can follow. Keep the task inside a test account first, with no live payments and no irreversible deletes, and only move to live data after the client has seen the test run and approved the steps. Computer use is powerful, but unsupervised computer use on live data is how a small job becomes a large incident. Third, a chart and report refresh package. A client gives you last month's report and this month's data, including charts and tables as images or files, and you hand back an updated report with refreshed charts, checked numbers, and a short note on what changed. This uses vision to read the old report, reasoning to compare it with new data, and computer use to update the files in place. Keep the scope to one report, with a clear page count and a clear revision limit, so the job does not grow into an open ended retainer by accident. What these three share is a boundary. The client knows what they get, you know how long it should take, and the token bill stays inside a range you can predict. A vague offer, such as unlimited AI automation for a month, breaks all three, and it will break your pricing too, because a computer use agent without a boundary can run for a very long time on a very large input.

What does the price actually buy, token by token?

The official list price is $0.834 per million input tokens and $2.501 per million output tokens, unchanged from the Hy4 preview. On TokenHub until 31 December 2026, the 50 percent discount gives calculated rates of $0.417 per million input tokens and $1.2505 per million output tokens. Those discounted figures are calculations, not separate quoted prices, and they apply on TokenHub only, not on every gateway. Third party gateways such as OpenRouter may charge their own rates, so check the gateway price before you quote. | Per 1 million tokens | List price | TokenHub until 31 December, calculated | What it means in practice | | Input | $0.834 | $0.417 | Text, images, and documents you feed the model | | Output | $2.501 | $1.2505 | Text the model writes, including summaries, tables, and action logs | | WorkBuddy and CodeBuddy, 4 to 15 November | Limited free quota, amount not stated | Limited free quota, amount not stated | Test window only, not a basis for client pricing | A worked example helps. Suppose a document package uses 5 million input tokens and 1 million output tokens. At list price, that is 5 times $0.834, plus 1 times $2.501, which is $4.17 plus $2.501, or $6.671 in total. At the TokenHub discount, the same package is 5 times $0.417, plus 1 times $1.2505, which is $2.085 plus $1.2505, or $3.3355 in total, about $3.34. The discount saves about $3.34 on that package, which is real, but it is small next to the package price, and that is the point. Token costs are not the constraint. Your time, and finding the client, are. Now test a larger job. Suppose a chart refresh package uses 20 million input tokens and 5 million output tokens, because it reads many charts and writes a long updated report. At list price, that is 20 times $0.834, plus 5 times $2.501, which is $16.68 plus $12.505, or $29.185 in total. At the TokenHub discount, it is 20 times $0.417, plus 5 times $1.2505, which is $8.34 plus $6.2525, or $14.5925 in total, about $14.59. Even the larger job stays under $30 in tokens at list price, which means a package priced at $149 can absorb a token price change, a longer input, or a few extra revisions without breaking. A service priced at $19 could not, which is why package price matters more than model price.

What does the honest maths look like?

Use the illustration below as a way to test your own numbers, not as a promise of income. Every figure is labelled, and you should replace each one with your own costs before you quote a client. The setup is a seller offering a document and computer task package at $149 per package. In the illustration, the seller sells two packages in a month, works about 6 hours per package including client calls, reading documents, testing the workflow, revisions, and handover, and keeps token spend inside the first example above, at about $6.67 per package at list price, or about $3.34 on TokenHub during the discount. The table uses the list price, so the maths still works after 31 December. | Line | Illustration | Notes | | Revenue, 2 packages at $149 | $298 | Illustration only, not a promise, clients are not guaranteed | | Token spend, 2 packages at about $6.67 at list price | About $13.34 | On Hy4 list rates, 5 million input plus 1 million output per package | | Other direct costs, such as tooling and payment fees | Not included in this illustration | Add your real fees, taxes, and tool subscriptions before quoting | | Time, 2 packages at 6 hours | 12 hours | Includes calls, reading, testing, revisions, and handover | | Revenue minus token spend, before time, tax, and fees | About $284.66 | $298 minus about $13.34, before tax, fees, and your time value | | Revenue per hour, before tax and fees | About $23.72 per hour | $284.66 divided by 12 hours, before tax, fees, and extra costs | Read that table carefully. The token bill is almost irrelevant at this price, which is the point of using a model that costs under $1 per million input tokens. What matters is whether you can find two clients, deliver two good packages in 12 hours, and get paid. If you cannot find clients, the token price does not save you. If you price too low, a cheap model does not save you either, because your time is still the main cost. Now test the other direction. If you raise the package to $249 and sell two in a month, revenue is $498, and revenue minus the same token spend is $484.66, or about $40.39 per hour before tax and fees on the same 12 hours. If you sell only one package at $149, revenue is $149, and revenue minus one token bill is $142.33, or about $23.72 per hour on 6 hours, but only $142.33 in total for the month. The model price barely moves those numbers. Client volume and package price do. That is why this article spends more words on scope and selling than on tokens: tokens are cheap, clients are not. Also test the discount risk. If you priced at the TokenHub discount, your token bill per package would be about $3.34 instead of $6.67, a saving of about $3.33 per package, or about $6.67 across two packages. That saving disappears on 1 January 2027, when the discount ends, but the illustration still works at list price, because it was built on list price from the start. Price for your time, not for the current discount, and treat the discount as a margin bonus during November and December, not as the business model.

How do you avoid the traps?

The first trap is trusting vision output without checking it. Hy4 can read charts and documents, but it can also misread a number, a label, or a trend, and a wrong number in a client report is worse than a slow report. Check every figure against the source document, keep the source page or file noted next to each figure, and have the client spot check one page before you deliver the full batch. That check takes minutes and prevents the most expensive failure in a document service: a confident error. The second trap is running computer use unsupervised on live data. Hy4 can operate graphical interfaces and command line tools, which means it can also delete, overwrite, or send the wrong thing if a step goes wrong. Always test in a test account or a copy of the data first, keep a backup before any live run, log every action the agent takes, and stay present for the first live run. A runbook that says what the agent will do, in order, with a stop condition, is not bureaucracy. It is the product the client is actually buying: a task done safely, not just quickly. The third trap is depending on the free quota or the discount for client work. The WorkBuddy and CodeBuddy free quota runs only from 4 to 15 November, and the exact amount is not published, so it is a test window, not a delivery budget. The TokenHub discount ends on 31 December. Use the free window to test, use the discount window to build margin, and quote every client job at list price, so January does not break your pricing. The fourth trap is assuming the weights or Hy4 Flash will rescue a bad plan. The full weights are not out yet, with no date published, and Hy4 Flash has no specs or date either. If your service needs self hosting, Hy4 is not ready for that today, and you should pick a model that actually ships weights, or run Hy4 through the hosted API and accept the gateway terms, data handling, and billing that come with it. Read the gateway terms as well as Tencent's, check where data is processed, and do not feed client secrets, private documents, or personal data into a service whose handling you have not checked. For sensitive client data, a local or private model may be the right choice even if it costs more per token. The fifth trap is overpromising on quality. Tencent claims Hy4 joins the top tier of open source models, but admits it lags top closed models in research level maths, long horizon coding, model self improvement, and advanced visual question answering. Test Hy4 on your own typical jobs, compare it with a model you already trust on the same job, and only sell work you have actually delivered and reviewed. A cheap model that produces unchecked output is not a bargain. It is a liability. For readers building a freelance offer around AI tools, our research agent service guide on DeepSeek Search and our agent service guide on Nebius credits use the same honest maths on different rates, and our short video service guide on AIVideoFree shows how a free credit service can be priced without promising income. Sources: Tencent Hy4 launch, 4 November 2026, as reported by TechNode, 4 November 2026, for $0.834 and $2.501 list prices, 50 percent TokenHub discount until 31 December, WorkBuddy and CodeBuddy free quota 4 to 15 November, and vision and computer use capabilities; ChainCatcher, 4 November 2026, for yuan pricing and discount wording; independent launch analysis, 4 November 2026, for weights not yet released and Hy4 Flash with no date, and Tencent's admitted gaps in research maths, long horizon coding, self improvement, and advanced visual question answering. TokenHub discounted figures are calculations from the stated list price and the stated 50 percent discount.

Frequently Asked Questions

What is Hy4?

It is Tencent's flagship Hunyuan model, launched on 4 November 2026, with 770 billion total parameters, visual understanding, and computer use through graphical interfaces and command line tools.

How much does Hy4 cost?

The standard API list price is $0.834 per million input tokens and $2.501 per million output tokens. On TokenHub until 31 December 2026, the 50 percent discount gives calculated rates of $0.417 input and $1.2505 output per million tokens.

What service should I sell with Hy4?

A narrow, repeatable package, such as a document reading and cleanup package, a small supervised computer task package, or a chart and report refresh package, priced by package with scope, revisions, and delivery format stated upfront.

Can I build the service on the free quota?

No, for client work. The WorkBuddy and CodeBuddy free quota runs only from 4 to 15 November 2026, and the exact amount is not published. Use it to test, and quote client jobs at list price.

Will selling two $149 packages a month make me a living?

No, on the illustration in this article, two packages give $298 in revenue and about $23.72 per hour before tax and fees, on 12 hours of work, after token spend at list price. Treat that as a starting test, not an income promise.

What happens to my pricing after 31 December?

The TokenHub discount ends, so TokenHub pricing returns to the published list price unless Tencent extends it. Because the illustration in this article uses list price, it still works after the discount ends, with the discount as a bonus during November and December only.