Access Chinese AI Models via One API

AIsa gives you a single, OpenAI-compatible API key to access China's most capable AI models — Qwen, DeepSeek, Kimi K2.5, ByteDance Seed, MiniMax-M2.5, and GLM-5 — through official partnerships with Alibaba Cloud, Moonshot AI, BytePlus, MiniMax, and Zhipu AI. No separate signups, no per-provider billing, no API key juggling.

Switch between a 1M-context Qwen flagship, DeepSeek's industry-leading price-performance, Kimi K2.5's 1-trillion-parameter reasoning, ByteDance Seed's capable general and agentic models, MiniMax-M2.5's 196K long-context processing, and GLM-5's strong Chinese-language reasoning — all through the same endpoint you already use for GPT, Claude, and Gemini.

Why access Chinese LLMs through AIsa?

Accessing Chinese AI models directly means navigating separate registration processes (some requiring Chinese phone numbers or entity verification), multiple billing accounts in different currencies, per-provider rate limits, and no fallback when a provider is down.

AIsa eliminates every one of those friction points:

AIsa is an Alibaba Cloud Qwen Key Account Partner, has a direct enterprise agreement with Moonshot AI (Kimi), routes ByteDance Seed models through BytePlus — ByteDance's official international API platform — and has enterprise agreements with MiniMax and Zhipu AI (GLM-5).

Supported Chinese AI models

Full pricing for all models is at marketplace.aisa.one/pricing.

Quickstart: call a Chinese AI model in 60 seconds

All Chinese models use the same endpoint and the same OpenAI-compatible schema. The only change from your existing code is the model string.

Python

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_AISA_API_KEY",
    base_url="https://api.aisa.one/v1"
)

response = client.chat.completions.create(
    model="qwen3.6-plus",        # swap to deepseek-v3.2, kimi-k2.5, seed-1-6-250915, etc.
    messages=[
        {"role": "user", "content": "Explain mixture-of-experts architecture in simple terms."}
    ]
)

print(response.choices[0].message.content)

Node.js

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.AISA_API_KEY,
  baseURL: "https://api.aisa.one/v1",
});

const response = await client.chat.completions.create({
  model: "kimi-k2.5",           // swap to any Chinese model string
  messages: [
    { role: "user", content: "Write a Python function to parse nested JSON." }
  ],
});

console.log(response.choices[0].message.content);

That's it. If you already call GPT or Claude through AIsa, you can access any Chinese model by changing one line.

Model strings

Use these exact strings in the model field:

qwen3.6-plus
qwen3-max
qwen3-coder-plus
qwen3-coder-flash
qwen3-coder-480b-a35b-instruct
deepseek-v3.2
kimi-k2.5
MiniMax-M2.5
GLM-5
seed-1-6-250915
seed-1-6-flash-250715
seed-1-8-251228
seedream-4-5-251128

Check marketplace.aisa.one/pricing for the complete and always-current list.

Enterprise data privacy

AIsa has formal enterprise agreements governing data handling for each Chinese model provider:

• Qwen (Alibaba Cloud): Processed under AIsa's Alibaba Cloud Key Account agreement. Data is not used for model training.
• Kimi K2.5 (Moonshot AI): Under a Supplemental Enterprise Service Agreement (effective February 10, 2026), customer data is not retained by Moonshot AI after processing, and generated outputs are not stored on Moonshot's infrastructure.
• ByteDance Seed (BytePlus): Routed via BytePlus, ByteDance's enterprise international API arm, subject to BytePlus enterprise data terms.
• DeepSeek: Accessed via Alibaba Bailian aggregation, inheriting AIsa's Alibaba Cloud enterprise data protections.
• MiniMax-M2.5 (MiniMax): Accessed under AIsa's enterprise agreement with MiniMax. Customer data is not used for model training.
• GLM-5 (Zhipu AI): Accessed under AIsa's enterprise agreement with Zhipu AI. Customer data is not used for model training.

If your organisation has specific compliance requirements, contact us for a custom data processing agreement.

Choosing the right model

Not sure which Chinese model to use? Here's a quick heuristic:

• Longest context (1M tokens): qwen3.6-plus — for book-length documents, huge codebases, or ultra-long conversations.
• Best price-performance: deepseek-v3.2 — frontier-adjacent quality at a fraction of Western model costs.
• Complex reasoning and agents: kimi-k2.5 — 1T parameter MoE, purpose-built for agentic tool-calling and visual coding.
• Code generation: qwen3-coder-480b-a35b-instruct — maximum coding capability with the full 480B MoE model; or qwen3-coder-plus for the best quality-per-dollar on everyday tasks.
• Chinese-language tasks: GLM-5 — Zhipu AI's flagship, with particular depth in Chinese reasoning, bilingual processing, and China-market applications.
• Long-document processing (mid-range context): MiniMax-M2.5 — 196K context, strong multilingual performance, efficient for large batch document workflows.
• High-volume throughput: seed-1-6-flash-250715 — optimised for low latency and minimum cost per token.
• Image generation: seedream-4-5-251128 — text-to-image up to 4K, with accurate typography rendering, via the images.generate endpoint.

What's next

• Qwen models — full guide
• DeepSeek models — full guide
• Kimi K2.5 — full guide
• ByteDance Seed & Seedream — full guide
• MiniMax-M2.5 — full guide
• GLM-5 — full guide
• All supported models
• Getting started with AIsa