All Top Chinese LLMs. One API.
Up to 90% cheaper.
DeepSeek, Qwen, GLM, Kimi, Doubao and more — behind a single OpenAI-compatible endpoint. Change your base URL and API key, keep everything else.
Pay as you go · No monthly fee · From $0.15 / 1M tokens
Two lines changed. You're in.
HeyToken speaks the OpenAI protocol. Point your existing SDK at our endpoint and everything just works.
from openai import OpenAI
client = OpenAI(
base_url="https://api.heytoken.io/v1", # ← change this
api_key="sk-your-heytoken-key", # ← and this
)
resp = client.chat.completions.create(
model="deepseek-v4-flash",
messages=[{"role": "user", "content": "Hello!"}],
)
print(resp.choices[0].message.content)Plug into your favorite AI agents
HeyToken speaks both the OpenAI and Anthropic wire protocols, so the coding agents and chat clients you already use work out of the box — just swap the base URL.
# Claude Code — point at HeyToken's Anthropic-compatible endpoint
export ANTHROPIC_BASE_URL=https://api.heytoken.io
export ANTHROPIC_API_KEY=sk-your-heytoken-key
claude # now backed by DeepSeek / GLM / Kimi…Same league. A fraction of the price.
Chinese frontier models match the performance of GPT-5, Claude and Gemini on major benchmarks — at a fraction of the cost.
| Model | Input / 1M tokens | Output / 1M tokens | Frontier equivalent |
|---|
Comparison prices are official list prices as of August 2026.
The best of Chinese AI, one key away
Frontier reasoning, million-token contexts, agent-grade coding — from the labs shipping the fastest.
kimi-k2.7-codedeepseek-v4-flashglm-5.3deepseek-v4-proglm-5.2Built for production workloads
OpenAI compatible
Drop-in replacement for the OpenAI API. Works with the official SDKs, LangChain, LlamaIndex, Cursor, Cline and more.
Pay as you go
No subscriptions, no seats, no minimums. Top up your wallet and pay only for the tokens you use.
High availability
Multi-channel upstream pooling with automatic failover keeps your requests flowing even when a provider degrades.
One unified bill
Every model, every key, every request — itemized in a single dashboard with per-token cost breakdowns.
Low latency worldwide
Optimized routes between global edge and Chinese inference clusters. Streaming from the first token.
Private by design
Your prompts are never used for training. Request logs are yours alone, exportable and deletable.
Up and running in 3 minutes
Create your account
Sign up with your email. No credit card required to look around.
Top up your wallet
Add funds from $5. New accounts get a free trial credit to test every model.
Change two lines
Swap the base URL and API key in your code. Ship it.
Frequently asked questions
Is HeyToken compatible with the OpenAI SDK?
Yes. HeyToken implements the OpenAI chat completions protocol. Set base_url to https://api.heytoken.io/v1 and use your HeyToken API key — no other changes needed. Streaming, function calling and JSON mode are supported.
Which models can I use?
We aggregate frontier Chinese models including DeepSeek, Alibaba Qwen, Zhipu GLM, Moonshot Kimi, ByteDance Doubao, MiniMax and Tencent Hunyuan. The full list with live pricing is on the Models page.
How does billing work?
Pure pay-as-you-go. You top up your wallet in USD, and we deduct per token based on each model's listed input/output price. No monthly fees, no expiry on your balance while the account is active.
How can I pay?
We accept major credit cards and Apple Pay through our payment partner. Crypto top-ups (USDT) are on the roadmap.
Is my data private?
We never use your prompts or completions for training. Logs exist only to provide your usage dashboard and can be exported or deleted on request.
Why is it so much cheaper?
Chinese labs price inference aggressively, and we pool volume across upstream channels. You get the same model weights at domestic economics, without needing a Chinese payment method or entity.