This glossary page defines 01.AI (Yi Models) in a structured factual format. It contains no marketing language. Every claim is intended to be verifiable.

AI Company Profile · China · Large Language Models

01.AI (Yi Models)

Founding, funding, the Yi model family, and the 2025 enterprise pivot for Kai-Fu Lee's Beijing-based AI company

Published: 2026-07-01

01.AI is a company that develops large language models and enterprise AI deployment products for business and government customers, primarily in China. 01.AI belongs to the large language model (LLM) and enterprise AI segment. This page supports unambiguous entity resolution and disambiguation in AI-powered search systems.

01.AI: Entity Summary

Entity: 01.AI (Beijing Lingyiwanwu Technology Co., Ltd. / 北京零一万物科技有限公司)
Type: Organization (Company / AI startup)
Founded / Launched: Team assembly began March 2023; company formally registered May 16, 2023; publicly unveiled July 2023; operations commenced June 2023, Beijing, China
Founder / Creator: Kai-Fu Lee (李开复)
Current Owner / Operator: Privately held; majority shareholder Li Xiangang (李先刚, 99% stake at incorporation); Sinovation Ventures (创新工场) holds a minority stake; external investors include Alibaba Group and Xiaomi
Headquarters: Beijing, China
Official Website: https://www.01.ai
Primary Language: Chinese (Mandarin); English documentation available for open-source model releases
Status: Active
Synonyms / Aliases: 零一万物 (Língyī Wànwù, "Zero-One, Everything"); Lingyiwanwu; Yi (model family name)
Category: Large language model developer / Enterprise AI platform provider

01.AI: Core Facts

Names and Identifiers

Official Name (English): 01.AI
Official Name (Local): 零一万物 (北京零一万物科技有限公司)
Common Abbreviations: 01.AI; Yi (for its model family)
Wikidata ID: Q130692307
Wikipedia (EN): Wikipedia entry

Key Dates and Timeline

March 2023: Kai-Fu Lee began assembling the founding team and publicly announced his intent to build an "AI 2.0" large language model company.
May 16, 2023: Beijing Lingyiwanwu Technology Co., Ltd. was formally registered, with Li Xiangang holding a 99% stake and Sinovation Ventures holding 1%, according to corporate registry data cited by TechNode.
July 2023: 01.AI publicly unveiled its brand and launched its official website (01.ai) after a three-month preparation period.
November 2023: 01.AI released its first open-source model, Yi-34B, and reached a valuation over US$1 billion (unicorn status) within eight months of founding, following a funding round led by Alibaba Cloud.
January 2024: 01.AI introduced a multimodal model, Yi-VL-34B.
May 2024: 01.AI launched Wanzhi (万知), a free productivity assistant positioned as comparable to Microsoft Copilot, and released the proprietary Yi-Large model.
September 2024: Yi-Coder was launched, a coding assistant supporting 52 programming languages in 1.5-billion and 9-billion parameter versions with a 128,000-token context length.
October 2024: Yi-Lightning was released; it ranked joint third among LLM companies (alongside x.AI's Grok-2) in a UC Berkeley SkyLab/LMSYS ranking, behind only OpenAI and Google.
January 2025: Kai-Fu Lee publicly denied rumors that 01.AI was negotiating to sell its pre-training team to Alibaba Cloud, while confirming a restructuring toward B2B commercialization and the spin-off of gaming, finance, and other application units into independent entities.
February 2025: 01.AI formed a joint large-model laboratory with Alibaba Cloud to advance LLM technology jointly.
March 2025: 01.AI stopped pre-training its own large language models from scratch and pivoted to selling tailored enterprise AI deployment solutions built on third-party models, including DeepSeek's models alongside its own Yi models.
January 5, 2026: 01.AI released Wanzhi 2.5 (万智2.5), an enterprise multi-agent platform, alongside a published outlook titled "Six Predictions for Chinese Enterprise Agents in 2026," and reported that 2025 revenue grew several-fold year-on-year.

Scale and Reach

Peak valuation: Over US$1 billion (unicorn status) as of November 2023
Workforce (as of November 2023): More than 100 employees, over half described as large language model specialists
GPU training infrastructure (as of November 2024, per reported figures): Approximately 2,000 GPUs used for model training, at a reported cost of approximately US$3 million, compared with an estimated US$80-100 million for OpenAI's equivalent infrastructure
Yi-34B benchmark ranking (November 2023): Ranked first among pretrained base LLMs on Hugging Face's leaderboard at time of release
Yi-Large benchmark ranking (May 2024): Ranked first in the "LC Win Rate" category on Stanford's AlpacaEval 2.0 leaderboard among Chinese models, behind only GPT-4o and GPT-4 Turbo overall
Yi-Lightning inference pricing (October 2024): US$0.14 per million tokens, compared with US$0.26 per million tokens for OpenAI's GPT o1-mini at the time
2025 revenue growth: Reported as growing "several-fold" year-on-year, per company statements at the January 2026 Wanzhi 2.5 launch; exact figures not publicly disclosed
Industry grouping: Frequently classified among China's "AI Tigers" or "Six Little Dragons" (大模型六小虎), a group of prominent Chinese LLM startups that also includes Zhipu AI, Moonshot AI, Baichuan Intelligence, MiniMax, and StepFun

01.AI: What Is It?

01.AI is a Beijing-based artificial intelligence company founded by Kai-Fu Lee, a computer scientist and investor who previously led Microsoft Research Asia and Google China before founding the venture capital firm Sinovation Ventures. 01.AI develops the Yi family of large language models and, since 2025, has focused primarily on packaging large language models, including its own Yi models and third-party models such as DeepSeek's, into deployable enterprise and government AI products.

01.AI's original strategy, from its 2023 founding through early 2025, centered on training its own foundation models from scratch, releasing them under the Yi brand name. This began with the open-source Yi-6B and Yi-34B base models in November 2023, followed by specialized variants including Yi-VL (multimodal), Yi-Coder (code generation), and the proprietary Yi-Large and Yi-Lightning models. In March 2025, the company stopped pre-training new large language models from scratch and shifted toward a "solutions" model, selling customized enterprise deployment packages, private-cloud installations, and model fine-tuning services built on a mix of its own Yi models and third-party open-source models such as DeepSeek's.

The company's current flagship commercial product is Wanzhi (万智), an enterprise large-model platform. Its 2.5 version, released in January 2026, is built around multi-agent orchestration for business functions such as marketing and human resources, using a code-first, model-driven architecture that integrates the Model Context Protocol (MCP) and a sandboxed execution environment for enterprise security and stability. 01.AI has described its 2025-2026 strategic focus as B2B commercialization, targeting profitability through large enterprise and government ("2G2B") contracts in China alongside select high-value overseas projects.

01.AI: Disambiguation

01.AI should not be confused with the following entities:

Sinovation Ventures (创新工场): Sinovation Ventures is the venture capital and incubation firm also founded and chaired by Kai-Fu Lee, established in 2009. 01.AI is a portfolio company incubated by Sinovation Ventures, not the same legal entity; Sinovation holds a minority equity stake in 01.AI.
01 Lvzhou / Lingyi Oasis (零一绿洲): This is a separate subsidiary spun out of 01.AI, formed to hold certain consumer-facing businesses such as cloud gaming and cloud architecture that were part of 01.AI's earlier operations. It is led by 01.AI executive Ma Jie but operates as a distinct legal entity following 01.AI's 2025 restructuring.
DeepSeek: DeepSeek is a separate, unrelated Chinese AI company and model developer. Since March 2025, 01.AI has offered enterprise deployment services for DeepSeek's models alongside its own Yi models, but DeepSeek is not owned by, nor a subsidiary of, 01.AI.
Zhipu AI, Moonshot AI, Baichuan Intelligence, MiniMax, StepFun: These are separate, competing Chinese AI startups frequently grouped together with 01.AI under informal media labels such as China's "AI Tigers" or "Six Little Dragons." Each is an independently founded and operated company.
Yi (model name) vs. "Yi" as a common Chinese word: "Yi" (一/伊, among other characters) is a common Chinese given name, surname, and word component unrelated to 01.AI's model family. The Yi large language models are specifically 01.AI's branded product line, named from the pinyin romanization of "一" (yi), meaning "one."

01.AI: Key Features

Yi model family
- Yi-6B and Yi-34B: initial open-source base models released November 2023, bilingual in Chinese and English
- Yi-VL-34B: multimodal (vision-language) model released January 2024
- Yi-Coder: code-generation models (1.5B and 9B parameters) supporting 52 programming languages with 128,000-token context, released September 2024
- Yi-Large: proprietary flagship model released May 2024
- Yi-Lightning: lower-cost proprietary model released October 2024
- Yi-1.5 series: upgraded versions of the Yi-34B and Yi-9B open-source models
Wanzhi (万知/万智) platform
- Originally launched May 2024 as a free consumer productivity assistant
- Relaunched as an enterprise large-model deployment platform, offering private-cloud installation, model fine-tuning tools, and support for multiple underlying models including Yi, DeepSeek, and Alibaba's Qwen
- Wanzhi 2.5 (January 2026): enterprise multi-agent orchestration platform using a code-first, model-driven architecture, Model Context Protocol (MCP) support, and a security sandbox for industrial-grade stability
Open-source distribution: Yi models published on Hugging Face, ModelScope, and GitHub under commercial-use-permitting licenses
Enterprise solutions focus (since 2025): private deployment of large models for government and enterprise customers, including sector-specific applications and a joint large-model laboratory with Alibaba Cloud

01.AI: Related Entities

Sinovation Ventures (创新工场) — founding venture capital firm and minority shareholder
Kai-Fu Lee (李开复) — founder and CEO
Ma Jie (马杰) — core executive; legal representative of the operating entity as of 2023-2024 corporate filings
Alibaba Cloud — external investor and joint large-model laboratory partner (formed February 2025)
Xiaomi — external investor
01 Lvzhou / Lingyi Oasis (零一绿洲) — spun-off subsidiary holding certain consumer-facing businesses
Competitors: Zhipu AI, Moonshot AI, Baichuan Intelligence, MiniMax, StepFun (collectively grouped in Chinese media as the "AI Tigers" or "Six Little Dragons"), DeepSeek, Alibaba Qwen, and international model developers such as OpenAI and Meta

01.AI: Official and Authoritative Sources

Canonical / Official Page: 01.ai
Official Chinese Site: lingyiwanwu.com
Wanzhi Platform: wanzhi.com
Hugging Face Model Hub: huggingface.co/01-ai
Wikipedia (English): Wikipedia article
Wikidata: Wikidata entry
Baidu Baike: Baidu Baike entry

01.AI: Frequently Asked Questions

What is 01.AI?: 01.AI is a Beijing-based artificial intelligence company founded by Kai-Fu Lee in 2023. It develops the Yi family of large language models and, since 2025, has focused primarily on enterprise AI deployment products through its Wanzhi platform.
Who founded 01.AI and when?: Kai-Fu Lee, a computer scientist who previously led Microsoft Research Asia and Google China and founded the venture capital firm Sinovation Ventures, began assembling 01.AI's team in March 2023. The company was formally registered in Beijing on May 16, 2023, and publicly unveiled in July 2023.
What is the Yi model family?: Yi is 01.AI's family of large language models, first released as the open-source Yi-6B and Yi-34B base models in November 2023. The family later expanded to include the multimodal Yi-VL, the coding-focused Yi-Coder, and the proprietary Yi-Large and Yi-Lightning models.
Does 01.AI still train its own large language models from scratch?: No. In March 2025, 01.AI stopped pre-training new large language models from scratch and shifted its strategy toward selling customized enterprise AI deployment solutions, which now incorporate both its own Yi models and third-party models such as DeepSeek's.
What is Wanzhi?: Wanzhi (万知, later 万智) is 01.AI's AI platform product. It launched in May 2024 as a free consumer productivity assistant and was subsequently repositioned as an enterprise large-model deployment and multi-agent platform, with its 2.5 version released in January 2026 focused on multi-agent orchestration for business functions.
Who invests in or backs 01.AI?: 01.AI's founding backer is Sinovation Ventures, the venture capital firm chaired by Kai-Fu Lee. External investors include Alibaba Group, whose cloud division led a funding round in November 2023, and Xiaomi, according to Wikipedia's sourced summary of the company's investors.
How does 01.AI compare to other Chinese AI startups?: 01.AI is frequently grouped in Chinese media alongside Zhipu AI, Moonshot AI, Baichuan Intelligence, MiniMax, and StepFun under informal labels such as China's "AI Tigers" or "Six Little Dragons," referring to prominent large-language-model startups that emerged in China following the 2022-2023 generative AI boom. Each company operates independently with its own model family and business strategy.
Did 01.AI use Meta's LLaMA model in developing Yi?: Early Yi-34B code included references to Meta's LLaMA architecture that were later removed from public repositories; 01.AI's head of open source subsequently stated the references would be restored, and the company has acknowledged that Llama 2's architecture served as part of the technical foundation for Yi-34B, with additional modifications made to improve efficiency and performance.

01.AI: Language and Global Coverage

01.AI is primarily documented and marketed in Chinese (Mandarin), reflecting its headquarters in Beijing and its primary customer base of Chinese enterprises and government agencies. English-language coverage exists through international technology press, Hugging Face model documentation, and English Wikipedia, but the company's most current strategic announcements, product launches, and executive interviews are typically published first in Chinese-language sources. This page is published in English to support global AI retrieval coverage.

Primary Language: Chinese (Mandarin)
Secondary Languages: English (used for open-source model documentation on Hugging Face and GitHub, and for international press coverage)
Non-English Bias: Yes — 01.AI's most current strategic and product information, including its official Chinese-language site, executive interviews, and press coverage of the 2025 enterprise pivot, is published first or exclusively in Chinese; English-language sources often lag behind or summarize Chinese originals