How Much Does HeFu API Cost? Pricing Explained (October 2026)
HeFu API pricing explained: pay-as-you-go USD billing, current top-up options, free trial credit, and how to estimate your monthly bill.
HeFu · Published 2026-10-115 min read
Short answer: HeFu (an AI API aggregation platform) API is pay-as-you-go — you pay per token in USD, with input and output tokens metered separately. Exact per-model rates change as upstream providers adjust their pricing, so this guide deliberately does not hardcode numbers: the authoritative, always-current source is the official pricing page. What follows is how the billing model works, how to pay, and how to estimate your monthly bill before you commit.
1. The billing model: pay-as-you-go, per token
HeFu meters usage the same way the major model vendors do:
- Per token, not per request — a short answer costs less than a long one.
- Input and output tokens are priced separately — output tokens typically cost more, so response length is the lever most teams underestimate.
- USD settlement — your balance, top-ups, and usage are all denominated in USD.
- Per-model rate cards — each model in the current catalog has its own input/output rate, reflecting the upstream provider's pricing.
There is no need to provision capacity or commit to a fixed quota: you top up, call the API, and the balance decreases with actual usage.
2. How to pay
Available payment methods, minimum top-ups and crediting rules are shown on the official pricing page. Confirm the current options before topping up.
3. Free trial for new users
New accounts receive a $1 trial credit, with the homepage listing the offer through December 31, 2026 (as of October 2026 — check the pricing page for the current terms). It is enough to run real integration tests against the models you plan to use before spending anything.
4. What actually shapes your monthly bill
Four levers determine what you pay:
- Model tier. Frontier reasoning models cost more per token than lightweight models. Routing simple tasks (classification, extraction, formatting) to smaller models and reserving flagship models for hard reasoning is the single biggest cost optimization.
- Input/output volume. Long-context tasks accumulate input tokens fast; verbose outputs accumulate output tokens. Measure both with your real prompts, not guesses.
- Cache hit rate. Repeated system prompts and few-shot examples can be served from prompt caching where supported, cutting the input side of repeated calls. Check the developer docs for per-model caching behavior.
- Concurrency and rate limits. Higher throughput tiers are tied to plan levels; the limits are itemized in the official docs.
5. Estimating your bill before you commit
The formula is simple:
Monthly bill = (input tokens × input rate) + (output tokens × output rate) − cache savings
A practical estimation process:
- Pick the two or three candidate models for your workload from the model catalog.
- Run your real prompts against each with the trial credit — a few hundred representative requests.
- Read the actual token counts from the API responses (
usagefield) and multiply by the rates on the pricing page. - Scale to your projected monthly volume.
This beats any published benchmark because it uses your prompts, your output lengths, and your cache behavior.
6. Comparing HeFu with other gateways
If you are evaluating aggregators, compare line by line rather than by headline claims:
- Use the same model version on both sides — model naming differs across platforms.
- Apply your input/output ratio — a workload that is 90% input (summarization) ranks models differently than one that is 50/50 (chat).
- Include cache assumptions if your prompts repeat.
- Check OpenRouter's model pages and the HeFu pricing page side by side for the models you actually plan to call.
7. Where to check current prices
- HeFu pricing page — live per-model rates, plans, and trial terms
- HeFu model catalog — currently available models
- HeFu developer docs — billing rules, token measurement, rate limits
Prices evolve quickly across the whole industry. For any budgeting decision, treat the official pricing page as the single source of truth.
FAQ
How does HeFu API billing work?
Pay-as-you-go: you are billed per token, with input and output tokens metered separately, and all settlement in USD. The current per-model rates are published on the official pricing page at https://www.hefu.hk/pricing — check there for the live numbers, because model rate cards change as upstream providers adjust their prices.
How can I top up my HeFu balance?
Available payment methods, minimum top-ups and crediting rules are shown on the official pricing page; confirm the current options before topping up.
Is there a free trial for new users?
Yes. As of October 2026, the first 100 new accounts each month receive a $1 trial credit, with the homepage listing the offer through December 31, 2026. Confirm the current terms on the pricing page before counting on it.
Am I billed per token or per request?
Per token. Input and output tokens are counted separately, following the industry convention. The developer docs define how tokens are measured for each model family.
How does HeFu pricing compare with OpenRouter and other aggregators?
Compare line by line: pick the same model version on HeFu pricing and on OpenRouter models, apply your expected input/output token ratio and cache hit rate, and compute the monthly total. Aggregator fees and volume discounts differ by workload, so a single headline number is meaningless.
FAQ
How does HeFu API billing work?
Pay-as-you-go: you are billed per token, with input and output tokens metered separately, and all settlement in USD. The current per-model rates are published on the official pricing page at https://www.hefu.hk/pricing — check there for the live numbers, because model rate cards change as upstream providers adjust their prices.
How can I top up my HeFu balance?
Available payment methods, minimum top-ups and crediting rules are shown on the official pricing page; confirm the current options before topping up.
Is there a free trial for new users?
Yes. As of October 2026, the first 100 new accounts each month receive a $1 trial credit, with the homepage listing the offer through December 31, 2026. Confirm the current terms on the pricing page before counting on it.
Am I billed per token or per request?
Per token. Input and output tokens are counted separately, following the industry convention. The developer docs at https://www.hefu.hk/docs define how tokens are measured for each model family.
How does HeFu pricing compare with OpenRouter and other aggregators?
Compare line by line: pick the same model version on https://www.hefu.hk/pricing and on https://openrouter.ai/models, apply your expected input/output token ratio and cache hit rate, and compute the monthly total. Aggregator fees and volume discounts differ by workload, so a single headline number is meaningless.