By Jonathan West · Updated Nov 17, 2026

Tencent Hy4 Flash Pricing per Million Tokens

Tencent's separate US dollar and Chinese yuan rate listings, with the TokenHub discount terms published as of November 17, 2026.

By Jonathan West · Updated Nov 17, 2026

Tencent lists Hy4 flash at US dollars (USD) 0.07 per million input tokens, USD 0.281 per million output tokens and USD 0.003 per million cache tokens, as listed on Tencent Cloud TokenHub. Tencent's Hy site separately lists Chinese yuan (CNY) 0.5 input, CNY 2 output and CNY 0.02 cached input per million tokens.

Tencent says TokenHub inference fees for Hy4 flash are settled at 50% of list price for eligible usage until 12:00:00 on November 30, 2026 (Beijing Time). The TokenHub page does not say whether its displayed USD rates are before or after that discount, so the displayed rates cannot establish a separate discounted USD price.


Tencent Hy4 Flash Pricing per Million Tokens

Tencent lists Hy4 flash at USD 0.07 input, USD 0.281 output and USD 0.003 cache per million tokens on TokenHub. Hy4 is listed at USD 0.834, USD 2.501 and USD 0.042 respectively. On Tencent's Hy4 flash announcement, the separate CNY listing gives Hy4 flash rates of CNY 0.5 input, CNY 2 output and CNY 0.02 cached input per million tokens. That announcement lists Hy4 at CNY 6 input, CNY 18 output and CNY 0.3 cached input.

The USD figures are as listed on the TokenHub page. Tencent's Hy announcement is in Chinese, while TokenHub publishes the USD figures and the dated discount notice on a separate page. Read the two listings in their stated currencies rather than treating either as a conversion of the other.

The TokenHub page displays region tabs for Singapore, Guangzhou and Silicon Valley. The listed figures have not been verified separately under each tab, so do not assume the same displayed rate applies to every region. The dated discount terms also matter when you estimate an invoice: Tencent says they apply to usage actually incurred within the offer period.


Rate Table

Tencent publishes four relevant rate rows across its TokenHub price page and Hy4 flash announcement. Each figure below is a listed price per million tokens in the currency shown. The table does not calculate a discounted price.

ModelCurrency and SourceInput per 1MOutput per 1MCache per 1M
Hy4 flashUSD, Tencent Cloud TokenHubUSD 0.07USD 0.281USD 0.003
Hy4USD, Tencent Cloud TokenHubUSD 0.834USD 2.501USD 0.042
Hy4 flashCNY, Tencent Hy siteCNY 0.5CNY 2CNY 0.02
Hy4CNY, Tencent Hy siteCNY 6CNY 18CNY 0.3

Tencent's Hy announcement presents its CNY columns in the order cached input, input and output. The table rearranges those same published figures into input, output and cache columns so each row can be read the same way. TokenHub labels its USD cache column "Cache (USD / million tokens)."

Keep the source with the currency when you copy a rate into a budget. For example, USD 0.003 is the Hy4 flash cache figure as listed on TokenHub. CNY 0.02 is the cache figure on Tencent's Hy site. Neither row supplies an exchange rate or establishes how the other listing was set.


The Limited-Time Discount

Tencent says Hy4 flash TokenHub inference fees are settled at 50% of list price only for usage incurred before the offer ends at 12:00:00 on November 30, 2026 (Beijing Time). Its TokenHub pricing notice states: "Hy4 flash Limited-Time Discount Notes: From the date of this notice until 12:00:00 on November 30, 2026 (Beijing Time), Hy4 flash model inference fees are settled at 50% of the list price. The discount applies only to usage actually incurred during the above period. After the period expires, the model fees will be settled at the list price again."

Hy4 has a different deadline. Tencent's notice states: "Hy4 Limited-Time Discount Notes: From the date of this notice until 23:59:59 on December 31, 2026 (Beijing Time), Hy4 model inference fees are settled at 50% of the list price. The discount applies only to usage actually incurred during the above period. After the period expires, the model fees will be settled at the list price again."

The two deadlines cannot be treated as one promotion end date. Hy4 flash's stated deadline is November 30 at noon Beijing time. Hy4's is December 31 at 23:59:59 Beijing time. Tencent ties each discount to when usage occurs, so scheduling a request after the relevant deadline does not meet the notice's stated condition.

TokenHub does not specify whether the USD figures displayed in its rate table are before or after the discount. It also gives no separate post-discount USD figure. Use the table for the published rate listing and the notices for the settlement terms. Do not multiply the displayed USD figures by 50% and label the result Tencent's charged rate.


How to Call It

Tencent Cloud TokenHub lists hy4-flash as the application programming interface (API) parameter for Hy4 flash. Its model pricing page lists hy4 for Hy4, so those names distinguish the two models when you select one for a TokenHub call.

Tencent says Hy4 flash is live on TokenHub and lists OpenRouter access as coming soon. Do not plan a current OpenRouter call on the strength of that announcement. Tencent also names WorkBuddy, Yuanbao, WeCom, QQ, ima and Marvis as places where Hy4 flash is available. That availability list does not give those products the TokenHub API price.

If you are estimating TokenHub usage, start with the hy4-flash model parameter and the matching USD row. Then account for Tencent's offer deadline and the region tab you intend to use. The pricing page shows three region tabs, but the published figures cited here have not been checked under each one.

Tencent says Hy4 flash has a one-million-token context length and is multimodal. Tencent recommends Hy4 flash for document processing and includes visual understanding in its WorkIndex evaluation. Its per-million-token figures are shown in the rate table. Tencent recommends Hy4 for complex, long-horizon software engineering tasks where it says Hy4 has a higher intelligence ceiling.


How It Compares

Tencent lists lower per-million-token input, output and cache prices for Hy4 flash than for Hy4 in both its USD and CNY listings. Tencent also reports lower costs per completed task on WorkIndex, its internal set of everyday WorkBuddy tasks. These are separate kinds of figures: the rate table prices tokens, while WorkIndex reports what Tencent says its tasks cost.

On WorkIndex, Tencent reports that Hy4 flash costs 1/13 as much per task as Hy4 preview, 1/20 as much as Hy4 and 1/8 as much as DeepSeek-V4.1-Flash-1017 at the max tier's busy-hour price. Those ratios come from Tencent's evaluation. They do not establish what a different task will cost, because the rates in the table alone do not specify how many input, cache or output tokens that task will use.

Tencent says WorkIndex combines internal task sets with human evaluation and considers output quality, single-turn output length and end-to-end time. Its tasks include slide work, document editing, spreadsheet repair, audio and video processing, and visual understanding. Tencent also reports that post-training cut end-to-end task output length by 50%. Treat that output-length figure as Tencent's reported result rather than a reduction to apply to your token budget.

For another rate card, use DeepSeek pricing or the cross-vendor AI model price list. Tencent's WorkIndex comparison names DeepSeek-V4.1-Flash-1017 at the max tier's busy-hour price, so match the DeepSeek rate to that tier and time of day before comparing it with Tencent's ratio.

Tencent says the Hy4 flash discount ends at 12:00:00 on November 30, 2026 (Beijing Time). If you need to estimate usage beyond that date, use the rate and terms then published by Tencent instead of carrying the limited-time settlement condition forward. The cost comparison would change if Tencent extends the offer, changes its list rates or publishes different rates under the region tab you use. The lower listed Hy4 flash rates and WorkIndex cost ratios are Tencent-published figures, while the discount remains subject to Tencent's dated notice. Before budgeting for Hy4 flash pricing, check the live Tencent Cloud TokenHub pricing page for the rate, region tab and discount terms.

Frequently Asked Questions

  • Tencent lists Hy4 flash at USD 0.07 input, USD 0.281 output and USD 0.003 cache per million tokens as listed on its TokenHub pricing page. Its Hy announcement separately lists CNY 0.5 input, CNY 2 output and CNY 0.02 cached input per million tokens. TokenHub also publishes a limited-time settlement discount. Its page does not identify whether the displayed USD rates are before or after that discount.
  • Tencent says WorkBuddy is free for a limited time until November 30, 2026, in its Hy4 flash announcement. That statement concerns WorkBuddy. The cited TokenHub pricing page does not publish a free API tier for Hy4 flash, so do not count WorkBuddy's offer as free TokenHub calls.
  • Tencent lists Hy4 flash cache at USD 0.003 per million tokens on TokenHub. Its separate Hy-site listing gives CNY 0.02 per million cached input tokens. Keep the currency and source attached to each figure rather than converting one listing into the other.
  • Yes. Tencent's TokenHub notice says Hy4 inference fees are settled at 50% of list price for usage actually incurred until 23:59:59 on December 31, 2026 (Beijing Time). Hy4 flash has a separate deadline of 12:00:00 on November 30, 2026 (Beijing Time). After each stated period, Tencent says fees return to list price.

Estimating Hy4 Flash API Costs?

Book a free 30-minute AI workflow audit with Layer3Labs to review token budgets, caching assumptions and model choices for your workflow.

Book a Consultation