01.AI (Lingyi Wanwu) Review: Pricing & Comparison

Kai-Fu Lee’s latest venture

01.AI (lingyiwanwu.com) founder Kai-Fu Lee needs little introduction — the former President of Google China and founder of Sinovation Ventures, with more than thirty years in the AI field.

Kai-Fu Lee founded 01.AI in 2023, focused on large language model R&D. Unlike many domestic AI startups, 01.AI has pursued a dual open-source track from the start — both open-source models in the Yi series (which can be self-hosted) and a commercial API (hosted inference).

The Yi series: multilingual capability as the moat

Where does the Yi series differentiate itself?

Multilingual capability: Yi models put relatively more investment into language support beyond Chinese and English. Comprehension and generation quality across Japanese, Korean, Arabic, Spanish, and other languages is relatively strong among domestic Chinese models.

For teams with internationalization needs — apps targeting Southeast Asian markets, multilingual content generation, cross-language translation and understanding — Yi’s multilingual capability is worth a serious look.

Open-source auditability: the open-source version of Yi means researchers can inspect the weights, fine-tune the model, and audit output bias. That’s valuable for users with specific compliance or research needs.

The logic behind the open-source/commercial dual track

Yi series models
├── Open-source (Yi-6B, Yi-34B, etc.)
│   ├── Apache 2.0 license
│   ├── Self-hostable
│   └── Downloadable directly from HuggingFace
└── Commercial API (Yi-Large, Yi-Medium, etc.)
    ├── Hosted inference by 01.AI
    ├── Mainland direct connect
    └── Pay-as-you-go

This dual-track structure gives users a clear migration path: validate with the commercial API first, and if you scale up, consider self-hosting the open-source weights instead.

How to connect

01.AI’s API is compatible with the OpenAI format:

from openai import OpenAI

client = OpenAI(
    api_key="your 01.AI API key",
    base_url="https://api.lingyiwanwu.com/v1"
)

response = client.chat.completions.create(
    model="yi-large",
    messages=[
        {"role": "user", "content": "Describe autumn in Kyoto in both Japanese and Chinese"}
    ]
)

Where it sits in the domestic AI landscape

Direct official APIs from domestic Chinese LLM vendors (ranked by registered user base):

Vendor	Flagship model	Differentiator
Alibaba Cloud Bailian	Qwen series	Enterprise SLA, ecosystem integration
Zhipu AI	GLM series	Multimodal, optimized for Chinese
Moonshot AI	Kimi series	Extra-long context
01.AI	Yi series	Multilingual, dual open-source track
DeepSeek	V4 series	Extremely low pricing, strong reasoning

01.AI’s competitive window is in multilingual scenarios. If your product doesn’t involve multiple languages, other options may offer better value.

Information verified 2026-07-04. The Yi series is continuously updated — defer to the official documentation at platform.lingyiwanwu.com.

Related reviews

• Banana AI: lightweight GPU inference, fast deployment for custom models, low cold-start time, developer-friendly
• DeepInfra: low-latency inference for open-source models, extremely transparent low pricing, supports crypto payment, includes image-generation models
• Poixe AI: operating since 2024, covers 7+ vendors, a top-up-credit membership discount system
• OpenClaw: a companion one-click deployment for Claude, low-latency direct connect exclusively for Claude Code users