Frontier LLMs. Exclusive video models. One API.
Zhen AI unifies every major LLM through one endpoint — and is the exclusive distributor of ByteDance's complete overseas Seedance video models, launching first in Japan. Aggregate, distribute, and bill by the token.
No credit card required · Exclusive video models · Deploy in minutes
Zhen AI is an OpenAI-compatible API gateway that serves frontier large language models from Anthropic, OpenAI, DeepSeek, Moonshot, Zhipu and other vendors through a single endpoint, https://api.zhenx.ai/v1. It is also the exclusive distributor of ByteDance's overseas Seedance video-generation models, available through the same account via POST https://www.zhenx.ai/v1/videos. Usage is billed per token for chat models and per second of output for video, with one API key, one balance and one invoice across every model.
Advanced GPU clusters. Open-source models at full speed.
Inference is served by advanced GPU clusters deployed in the United States — running the strongest open-source models at production throughput, billed by the token.
US-deployed GPU clusters
Inference runs on advanced GPU clusters deployed in the United States, with service infrastructure located outside mainland China.
Open-source frontier models
DeepSeek, GLM, Kimi, gpt-oss and more — the strongest open models, served at production speed through one endpoint.
Measured, not promised
The tok/s figures in the window above are live platform measurements, refreshed every minute — not lab numbers.
Chat with any model, right now
Every message is routed through the live Zhen AI gateway. Switch models mid-conversation — no account, no key, no setup.
Generate a video, right here
A limited-time MiniMax-H3 text-to-video playground running on a dedicated GPU node — one prompt in, a 720p clip with native audio out. No account needed.
One fixed best-value spec — 243 frames at 576p. A render takes about 3–4 minutes on the GPU.
Free limited preview · one clip at a time per visitor
Recent clips
A clip takes about 3–4 minutes. Feel free to leave — it keeps rendering, and the result appears below.
Recent clips
Only the latest 6 clips are kept — download yours as soon as it's ready.
Backed by exclusive partners, powered by every provider
Strategic partnerships that bring frontier video to the world — on top of a gateway that unifies every major model provider.
Unifying the world's leading model providers
Everything you need to run an AI business
A production-grade engine for routing, metering and governing model traffic across every tenant.
Millisecond routing
A Go-native edge proxy adds 1–7 ms of gateway overhead per request. Weighted load-balancing across every upstream key.
Multi-tenant by design
Org → Team → Virtual Key hierarchy. Per-tenant quotas, model allow-lists, and isolated billing out of the box.
Usage-based billing
Real-time token metering with Redis pre-deduction. Stripe subscriptions, invoices, and spend caps included.
Live observability
RPM / TPM dashboards, per-request logs, latency heatmaps. ClickHouse-backed analytics at any scale.
Resilient upstreams
Automatic failover, circuit-breaking and retry. A 429 on one channel transparently reroutes to the next.
One unified API
OpenAI, Anthropic and Gemini wire-formats from a single endpoint. Swap models without touching your code.
Drop-in compatible. Switch models in one line.
Point your existing SDK at Zhen AI. Route to any provider, fail over automatically, and meter every token — no code changes.
curl https://api.zhenx.ai/v1/chat/completions \
-H "Authorization: Bearer sk-..." \
-H "Content-Type: application/json" \
-d '{
"model": "glm-5.3",
"messages": [{"role":"user","content":"Hi"}],
"stream": true
}'Prices are adjusted from time to time. The figures on this page are the current ones. Last checked (UTC)
Pricing that scales with you
Start free. Upgrade when your traffic does. Cancel anytime.
Learn more about Zhen AI
Who operates the platform, common questions, and how we handle unrestricted-content models.
Zhen AI is a unified gateway that aggregates frontier language and video models from every major upstream provider, distributes access to your tenants with per-key quotas, and meters usage down to the token — all on infrastructure engineered for scale and reliability.
Zhen AI is a brand of Hangzhou Zhen AI Technology Co., Ltd. (杭州真AI科技有限公司), and the service is operated on its behalf by Eigenstate Research LLC (USA). In Japan, Terabox K.K. (Terabox株式会社) is our strategic partner, leading local launch and distribution. Hangzhou Zhen AI Technology has entered an exclusive partnership with ByteDance to bring the complete, unrestricted overseas edition of Seedance and other video-generation models to the Zhen AI gateway.
What is Zhen AI?
Zhen AI is a unified API gateway that aggregates frontier language and video models from every major upstream provider, distributes access to your tenants, and meters usage down to the token.
How do I get an API key?
Create a free account, sign in to the console, and generate a key from the Keys page. See the Docs page for a full walkthrough with a working curl example.
Which models are available?
The marketplace includes OpenAI, Anthropic, Google, xAI, Mistral and DeepSeek text models, plus ByteDance's Seedance video-generation models via our exclusive overseas partnership. See the Marketing page for upcoming Seedance releases.
Is there a free tier?
Yes — new accounts start with a trial quota you can use immediately. Upgrade to a paid plan for higher limits, priority routing, and an SLA.
Are the unrestricted Seedance models available everywhere?
No. The complete, unrestricted overseas edition of Seedance is only available in supported regions outside mainland China, subject to the Usage Policy and local law. See the Marketing page for the current regional rollout, starting in Japan.
How do I get support?
Reach out from the console, or contact our Japan partner Terabox K.K. directly for enterprise and localization inquiries.
- Use of the Zhen AI gateway and any connected model must comply with the acceptable-use terms of the underlying upstream provider in addition to our own.
- Unrestricted-content models (including the overseas edition of Seedance) are made available only in supported regions outside mainland China, in line with each region's local law.
- Access to unrestricted-content models requires that you are at least 18 years old and legally permitted to view such content in your jurisdiction.
- All requests are logged for billing, abuse-prevention and security purposes in line with our Privacy policy.
- We reserve the right to suspend keys or tenants found generating illegal content or violating an upstream provider's policy.
- Full Terms of Service and Privacy details are available from the console footer on request.