AI API Gateway · Hong Kong Nodes · USD Pay as You Go

Every major LLM. One API key. No borders.

HeFu gives you GPT, Claude and Gemini plus first-class access to Chinese models — DeepSeek, Qwen, Kimi, GLM, MiniMax — without a Chinese phone number or local billing. One OpenAI-compatible endpoint, transparent per-token pricing.

Sign Up FreeBook an Enterprise Demo

Why developers choose HeFu

Built for teams that want the whole global model market, not one vendor's slice of it.

Chinese models, unlocked

First-class DeepSeek, Qwen, Kimi, GLM and MiniMax access — no Chinese phone number, no local billing, no separate vendor accounts.

No hidden markup

Pay per token at transparent rates — several models are priced below official list thanks to volume purchasing. No platform fee on top-ups.

Pay your way

USD billing via Stripe, plus GrabPay, PayNow and FPX for Southeast Asia. Pay as you go — no subscriptions, no monthly minimums.

Smart routing built in

Priority routing, automatic failover and cross-model fallback chains keep your app up when a single provider degrades.

Flagship quality at a fraction of the cost

List-price arithmetic as of Aug 2026: the DeepSeek V4 series is 80–98% cheaper than GPT-5.5-class models, and capable coding models like MiniMax M3 start at $0.60 per million input tokens. Read the full pricing comparison →

Model lineup

Mix and match the right model for each scenario.

OpenAI GPT Series

Full GPT-5.6 lineup (Terra / Sol / Luna) plus the GPT-5.3 Codex coding model — Hong Kong nodes, no overseas credit card required.

Pay as you go

Claude Series

Claude Opus 5 / Fable 5 flagships and the cost-effective Sonnet 4.6 — a top pick for coding and long-form writing.

Pay as you go

DeepSeek

DeepSeek-V4-Pro / V4-Flash reasoning models at a fraction of GPT flagship cost, with an OpenAI-compatible API.

Pay as you go

Kimi (Moonshot AI)

Long-context leader Kimi K2.5 / K2.6 / K3 — the go-to choice for Chinese-language workloads.

Pay as you go

Google Gemini Series

Gemini 3.6 Flash / 3.5 / 3.1 Pro multimodal models — text, image, audio and video through one API.

Pay as you go

Chinese LLM Matrix

Qwen3.5-3.7, GLM-5.x, Doubao Seed 2.x, Grok 4.3 and MiniMax M3 — all behind one endpoint.

Pay as you go

Up and running in three steps

From sign-up to your first API call in minutes.

1

Sign up & top up in USD

One account unlocks every model series. Pay as you go, no monthly minimum.

2

Get your API key

Create a key in the console and track usage and spend in real time.

3

Switch the baseURL

OpenAI-compatible endpoints — point your existing code at HeFu and go.

How is HeFu different from OpenRouter?

HeFu focuses on first-class access to Chinese models through official partnerships rather than resale, bills in USD with Southeast Asia local payment options (GrabPay / PayNow / FPX), and charges no platform fee on pay-as-you-go usage. See our detailed OpenRouter alternative comparison on the blog.

Do I need a Chinese phone number or Chinese bank card?

No. Sign up with an email, top up in USD by card or SEA local payment methods, and call every model in the catalog — including DeepSeek, Qwen, Kimi, GLM and MiniMax.

Is my data sent to China?

Requests are routed per model: calls to Chinese models go to the respective official upstream (exactly as calling that vendor directly), while GPT / Claude / Gemini traffic goes to their official endpoints. Our privacy policy covers the details, and dedicated deployment is available for enterprise teams.

Is the API compatible with the OpenAI SDK?

Yes. In most cases you only change the baseURL and API key; switching models is a single parameter change. We publish setup guides for LobeChat, Cline, Open WebUI and other popular clients.

What about latency from the US or Europe?

Hong Kong nodes with multi-region intelligent routing give Asia-Pacific developers the lowest cross-border latency; US and EU teams should compare measured latency on our endpoint against direct access — live status is public on the main site.

From the blog

LLM API selection insights and integration practice. View all →

HeFu vs OpenRouter: Which Is Better? A Comprehensive Comparison

HeFu vs OpenRouter: which is better: a practical guide from HeFu.

Helium 10 Alternative for Amazon Listing Generation (2026 Guide)

Helium 10 alternative for Amazon listing generation: a practical guide from HeFu.

Free AI Listing Generator for Amazon Sellers: What Actually Works in 2026

free AI listing generator for Amazon sellers: a practical guide from HeFu.

Start building with one unified API

Self-serve pay-as-you-go for individuals, dedicated demos for enterprise teams.

Sign Up FreeBook an Enterprise Demo