Enterprise-grade Chinese AI models available globally. Qwen3.8, DeepSeek-V4, GLM-5.3, Kimi K3, HappyHorse, and more — through a single OpenAI-compatible endpoint. Pay with USDT.
No new SDK. Works for Qwen / DeepSeek / GLM via ChinaModelAPI. Full walkthrough on the guide page.
from openai import OpenAI
client = OpenAI(
api_key="YOUR_KEY",
base_url="https://api.chinamodelapi.com/v1",
)
r = client.chat.completions.create(
model="qwen-plus",
messages=[{"role":"user","content":"Hello"}],
)
print(r.choices[0].message.content)Enterprise-grade models from China's top AI labs, accessible through a unified OpenAI-compatible API.
Alibaba's 2026 flagship line. Qwen3.8-Max (~2.4T-class) targets coding, work, and long-horizon tasks. Production stacks still use Qwen-Max / Plus / 3.5-era endpoints where available. Not a Seedance-style video generator — see the Model Watch brief for VL name hygiene.
DeepSeek's 2026 flagship family: V4-Pro (~1.6T MoE / 49B active, 1M context) for agentic coding & reasoning; V4-Flash for cheap high-volume. The V4-Pro 正式版 (GA) landed 2026-08-13 as DeepSeek-V4-Pro-0813 — same model ID, stronger agents. Legacy R1 remains useful for pure reasoning workloads.
Zhipu's current flagship (released Aug 14, 2026): Terminal-Bench 3.0 4.6→28.3, emergent cyber (CyberGym 84.5%), solid 1M-token context; Z.ai GA'd the 5.3 API Aug 19 at $1.40/$4.40 per-1M — glm-5.3 relay ID follows health-check (legacy glm-5.2 / glm-5.1 IDs still resolve). Brand-new: GLM-5.3-Flash — Bloomberg-confirmed Aug 26 as the model behind viral stealth hit Ox Alpha, released with open weights ("frontier intelligence, flash cost"); flagship 5.3 weights expected ~Aug 28.
Moonshot's latest flagship (July 2026). ~2.8T parameters, 1M-token context, native vision — positioned as an open 3T-class model for long-horizon coding, knowledge work, and agents. Use K3 for frontier long-context; keep K2.x for lighter/cost-sensitive loads.
Alibaba's AI video generation model (快乐马). Produces cinematic-quality clips from text prompts and images — enterprise-grade video creation accessible via API. HappyHorse 1.1 is now officially live on Model Studio: video-editing mode, integrated audio, No.2 video-arena rank.
Seedance 2.0 is the ChinaModelAPI production default (multimodal text/image/audio/video). Seedance 2.5 was officially launched 2026-07-31 (~30s single-pass, ~50 multimodal refs) and its developer API went live on Volcano Ark 2026-08-07 — relay model ID pending.
Kuaishou's flagship video family — Kling 3.0 / 3.0 Turbo / 3.0 Omni: full-modal input, up to 15s multi-shot, native audio-video sync, industry-first native 4K direct output (Apr 2026). Official API from $0.084/s. Q2 revenue >RMB 850M (+200% YoY); July ~$3B raise at ~$18B valuation. Relay model ID pending health-check.
MiniMax's general-purpose multimodal video model (not the M3 text LLM): native stereo audio, up to ~15s / 2K on hosted paths, Hugging Face weights + ComfyUI. Community often compares quality to Seedance — not an official “open Kling” product name. Relay model ID pending health-check.
One-time token purchases. No subscriptions. Pay with USDT or USD1. Use tokens across all models.
All prices are in USD. USDT and USD1 payments accepted. Token balances never expire.
Everything you need to integrate Chinese AI models into your products.
Drop-in replacement for OpenAI SDK. Switch with a single line of code — no refactoring needed.
Pay with stablecoins. No credit card required. Fast settlement, available worldwide.
Monitor token usage, latency, and spend across all models in one dashboard.
Automatic PII masking. Your data never leaves our secure pipeline. SOC 2 compliant infrastructure.
Low-latency access from anywhere in the world. Optimized routing to Chinese model providers.
Dedicated support team across time zones. English and Chinese support available around the clock.
Already using OpenAI? Just change the base URL and API key. That's it.
from openai import OpenAI
# Just change the base URL — everything else stays the same
client = OpenAI(
base_url="https://api.chinamodelapi.com/v1",
api_key="your-api-key"
)
response = client.chat.completions.create(
model="qwen-3",
messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)From signup to your first API call in under 5 minutes.
Sign up for free in seconds. No credit card required.
Generate an API key from your dashboard. Purchase tokens with USDT.
Send requests to any supported model. Monitor usage from the dashboard.
Join our waitlist for exclusive early access, founding-member pricing, and free trial tokens.
No spam. Unsubscribe anytime. By signing up you agree to our Terms of Service.
We specialize exclusively in Chinese AI models — deeper coverage, better pricing, dedicated expertise.
We maintain direct agreements with Alibaba (Qwen), DeepSeek, Zhipu (GLM-5.2/5.3), Moonshot (Kimi K3), HappyHorse, and ByteDance — meaning more tokens per dollar, no middleman markup.
Access HappyHorse and ByteDance Seedance 2.0 video generation — models not widely available on general-purpose API platforms.
Serve developers worldwide with no geographic blocklists. Pay with standard USDT — no ecosystem token lock-ups required.
We're not a general-purpose aggregator. We focus 100% on Chinese AI — which means deeper knowledge, faster support, and better model recommendations.
Everything you need to know about accessing Chinese AI models.
https://api.chinamodelapi.com/v1. No special configuration or Chinese payment methods needed — we accept USDT stablecoins worldwide.
base_url to our endpoint. All chat completions, streaming, function calling, and other features work identically.
glm-5.2 and glm-5.1 today (confirm live aliases in the dashboard); glm-5.3 follows health-check after Z.ai's Aug 19 API GA, priced identically at $1.40/$4.40 per 1M tokens. Zhipu has also released GLM-5.3-Flash — confirmed Aug 26 via Bloomberg as the viral stealth model Ox Alpha — with open weights out now. Guide: GLM-5.3 API.
kimi-k3; earlier family members include K2.7 Code / K2.x for lighter loads. Prefer K3 when you need max long-context + agent quality; use cheaper models for short chat. Guide: Kimi K3 API.
happyhorse-1.0 (relay baseline) and happyhorse-1.1 — officially live on Alibaba Cloud Model Studio (Aug 2026): No.2 video-arena rank, new video-editing mode, integrated audio; relay routing to follow. See the full HappyHorse API guide for integration details.
seedance-2.0. See the full Seedance 2.0 API guide for code examples.