Tencent TokenHub lists HY4 Flash at ¥0.50 per million uncached input tokens, ¥2 per million output tokens, and ¥0.02 per million cache-hit tokens. Its current 50% promotion applies only to usage before November 30, 2026 at 12:00 Beijing time. After that, the tariff says rates return to list price. This page estimates TokenHub API usage in CNY; it does not price a WorkBuddy subscription or prove how much a completed task will cost.
Tencent's launch announcement says WorkBuddy access is free through November 30. That is separate from the API promotion. For task evidence, see the HY4 Flash review; for a model comparison, see HY4 Flash vs DeepSeek V4.1 Flash.
Key takeaways
TokenHub list rates are ¥0.50/M input, ¥2/M output, and ¥0.02/M cache hit.
The 50% promotion ends at noon Beijing time on November 30, 2026.
Promo rates are ¥0.25/M, ¥1/M, and ¥0.01/M.
API estimates do not equal a WorkBuddy subscription price or cost per successful task.
HY4 Flash TokenHub prices at a glance
| Usage type | List / 1M tokens | Promo / 1M tokens | Cutoff |
|---|---|---|---|
| Uncached input | ¥0.50 | ¥0.25 | Nov 30, 2026, 12:00 Beijing |
| Output | ¥2.00 | ¥1.00 | Nov 30, 2026, 12:00 Beijing |
| Cache-hit input | ¥0.02 | ¥0.01 | Nov 30, 2026, 12:00 Beijing |
Tencent's TokenHub tariff, updated November 16, is the source for CNY rates and timing. The separate English model-detail page displays USD rates and campaign prices; do not combine currencies or providers in one estimate.
Calculate token cost, not visible-text cost
Use the usage record returned by the API. If a request includes uncached input, cache-hit input, and output:
`text API cost = (uncached input × input rate
cache-hit input × cache-hit rate
output × output rate) ÷ 1,000,000 `
At list price, a hypothetical request with 10,000 uncached input tokens and 2,000 output tokens costs:
text
(10,000 × ¥0.50 + 2,000 × ¥2) ÷ 1,000,000 = ¥0.009
This excludes retries, tools, other providers, and application fees. At promotion rates, the same token counts cost ¥0.0045.
Local calculator
Estimate Hy4 Flash API costs
Enter billed tokens per request. Values stay in your browser.
50% off through Nov 30, 2026 at 12:00 (UTC+8)
List price
TokenHub pay-as-you-go prices in CNY. Include billed reasoning in output. Taxes, tools and Token Plan subscriptions are excluded. Tencent Cloud model pricing
The calculator uses Tencent TokenHub CNY token rates. It does not convert currency or estimate WorkBuddy usage. Read the HY4 Flash review for model evidence.
Cache hits are a separate billing line
Tencent lists cached input at ¥0.02 per million tokens, below its ¥0.50 uncached-input rate. The cache rate applies to reported cache-hit tokens; do not assume a repeated prompt is cached. Check usage fields. The tariff does not list a separate HY4 Flash cache-write rate.
A low cache price can matter for repeated long prefixes, but it does not describe the total cost of an agent run. Output, uncached input, calls, and retries still count. A Reddit user comparing one WorkBuddy calculation across three models noted that published rates were “unit-price comparisons, not measured savings from this Workbuddy session.” Read the original post.
Three prices that should not be mixed
| Price or claim | What it covers | What it does not establish |
|---|---|---|
| TokenHub API tariff | Per-token Tencent TokenHub use | A subscription or another provider's price |
| WorkBuddy free window | Tencent says free access through Nov 30 | That API use is free or has no product limits |
| Tencent WorkIndex task result | Tencent reports 1/8 DeepSeek's task cost at Max tier and peak pricing | A universal 1/8 bill for your prompts |
Tencent's current tariff should be checked before budgeting after the promo. Compare same-provider rates in HY4 Flash vs HY4 and HY4 Flash vs DeepSeek V4.1 Flash.
Estimate cost per successful task separately
Suppose a task makes four calls and uses 40,000 uncached input, 120,000 cache-hit input, and 8,000 output tokens. At list price:
text
(40,000 × ¥0.50 + 120,000 × ¥0.02 + 8,000 × ¥2) ÷ 1,000,000 = ¥0.0384
This is an arithmetic example, not a measured HY4 Flash task. Retries increase the bill. Estimate cost per success with:
text
cost per successful task = total API cost, including failed attempts ÷ successful tasks
Track provider, model ID, input/output/cache tokens, tools, retries, and task acceptance. In the Reddit comparison, all three models reportedly passed one arithmetic task, but the author did not share token use or the actual bill.
Is HY4 Flash cheap enough for your workload?
At TokenHub list rates, the unit price is low. That makes it worth checking for repetitive, bounded work where output can be validated. It does not prove the model uses fewer tokens or retries than another model. Plotly's independent DataAnalyticsBench report reported $0.13 for one run, but used its own harness and token mix.
See HY4 Flash model resources for setup guidance. A model API rate does not provide browser access: see Tencent Hy4 with a signed-in browser. Other routes include DeepSeek V4.1 Flash, its browser workflow, and Claude Haiku 5.5 browser use. Tabbit CLI setup and the browser-agent first-task checklist cover workflow setup, not HY4 Flash billing.
Verdict
Use the tariff to budget API tokens and actual logs to measure task cost. Keep the promotion cutoff visible in estimates. The rate card, WorkBuddy access, and task success are separate questions and need separate evidence.
Tabbit Browser offers Hy4 Flash. Download Tabbit, then select Hy4 Flash from the model list in Chat to start using it.
FAQ
How much does HY4 Flash cost on Tencent TokenHub?
Tencent lists ¥0.50 per million uncached input tokens, ¥2 per million output tokens, and ¥0.02 per million cache-hit tokens. A 50% discount applies to usage through November 30, 2026 at 12:00 Beijing time.
What is the HY4 Flash promotional price?
During the stated TokenHub promotion, rates are ¥0.25 per million input tokens, ¥1 per million output tokens, and ¥0.01 per million cache-hit tokens. The discount applies only to usage incurred before the published cutoff.
Does cached input cost the same as uncached input?
No. Tencent lists cache-hit input at ¥0.02 per million tokens, compared with ¥0.50 for uncached input. The tariff does not list a separate cache-write rate.
Is HY4 Flash free in WorkBuddy?
Tencent's November 16 launch announcement says HY4 Flash is free in WorkBuddy through November 30, 2026. WorkBuddy access and limits are separate from TokenHub API billing.
Does a low token price guarantee lower task cost?
No. A task may use different input, output, cached tokens, tool calls, and retries. Calculate API usage from actual token counts, then divide total usage costs by successful tasks if you need cost per success.