# 01.AI (零一万物)
**01.AI** (零一万物) is a Beijing-based artificial intelligence company founded in May 2023 under the leadership of Dr. [Kai-Fu Lee](https://www.edgepedia.org/kai-fu-lee), known for its open-source Yi series of large language models and, more recently, for enterprise AI transformation products and services. The company became a unicorn within six months of founding, reaching a $1 billion valuation in November 2023.[1](https://www.01.ai/) • [2](https://tech.yahoo.com/ai/articles/chinese-ai-startup-01-ai-092630274.html) Its API platform is operated by Beijing Lingyi Wanwu Technology Limited.[3](https://platform.01.ai/useragreement)
| Key fact | Detail |
|---|---|
| Founded | Beijing, May 2023, under Dr. Kai-Fu Lee; unicorn within six months[1](https://www.01.ai/) |
| Valuation | $1 billion reached in November 2023; in December 2023 the company was reported to be raising up to $200 million[2](https://tech.yahoo.com/ai/articles/chinese-ai-startup-01-ai-092630274.html) |
| Flagship open models | Yi family based on 6B and 34B pretrained language models, extended to chat, 200K long-context, depth-upscaled, and vision-language variants[4](https://arxiv.org/abs/2403.04652) |
| Pretraining data | 3.1 trillion tokens of English and Chinese corpora, built with a cascaded deduplication and quality-filtering pipeline[4](https://arxiv.org/abs/2403.04652) |
| License | Yi code and weights distributed under Apache 2.0, free for personal, academic, and commercial use[5](https://github.com/01-ai/yi) |
| Yi-Lightning | 100B-parameter Mixture-of-Experts model released October 2024, ranked sixth globally and first in China on the LMSYS blind test[1](https://www.01.ai/) |
| Enterprise products | WorldWise (Wanzhi) Enterprise LLM Platform (2025) and TrueNorth AI decision hub platform (2026)[1](https://www.01.ai/) |
## Founding and funding
Kai-Fu Lee, a computer scientist formerly a top executive at Apple and Microsoft and head of Google in China, launched the company in July 2023 after a three-month incubation period, according to Reuters; the company itself dates its founding to May 2023 in Beijing.[6](https://finance.yahoo.com/news/chinese-ai-startup-01-ai-092104117.html) • [1](https://www.01.ai/) Lee described large language model technology as "a historic opportunity that China cannot miss" when starting the venture.[7](https://qa.time.com/6310603/kai-fu-lee/)
Funding came quickly. Lee said the company was valued at $1 billion after receiving financing from [Sinovation Ventures](https://www.edgepedia.org/sinovation-ventures), Alibaba Cloud, and other undisclosed investors.[8](https://techcrunch.com/2023/11/05/valued-at-1b-kai-fu-lees-llm-startup-unveils-open-source-model/) In December 2023, Reuters reported that 01.AI was raising up to $200 million in fresh capital on that valuation, with AliCloud having committed funding in the round.[2](https://tech.yahoo.com/ai/articles/chinese-ai-startup-01-ai-092630274.html) • [6](https://finance.yahoo.com/news/chinese-ai-startup-01-ai-092104117.html) [TechCrunch](https://www.edgepedia.org/techcrunch) reported that the company had proactively stockpiled GPUs in anticipation of United States sanctions, borrowing money to buy processors before landing funding.[8](https://techcrunch.com/2023/11/05/valued-at-1b-kai-fu-lees-llm-startup-unveils-open-source-model/)
## The Yi model family
The Yi series is a set of open-source large language models trained from scratch by 01.AI. The company states that it built its own training datasets, training pipelines, and training infrastructure independently, and that because the Llama structure is employed by most open-source models, performance is determined mainly by data, pipelines, and infrastructure.[5](https://github.com/01-ai/yi)
The March 2024 technical report describes the family as based on 6B and 34B pretrained language models, extended to chat models, 200K long-context models, depth-upscaled models, and vision-language models. The authors attribute performance primarily to data quality from data-engineering efforts; for fine-tuning they polished a small instruction dataset of fewer than 10,000 instances, each verified directly by machine learning engineers. Vision-language models combine the chat language model with a vision transformer encoder, and context length was extended to 200K tokens through lightweight continual pretraining.[4](https://arxiv.org/abs/2403.04652)
**Benchmark standing.** In November 2023, the Yi-34B base model ranked first among all existing open-source models, including Falcon-180B, Llama-70B, and Claude, in both English and Chinese on benchmarks including the Hugging Face Open LLM Leaderboard (pre-trained) and C-Eval, based on data available up to that date.[5](https://github.com/01-ai/yi) • [8](https://techcrunch.com/2023/11/05/valued-at-1b-kai-fu-lees-llm-startup-unveils-open-source-model/) In December 2023, Yi series models ranked just behind GPT-4 and above Llama on the Alpaca Leaderboard.[5](https://github.com/01-ai/yi) The Yi-34B-Chat model ranked first among open-source models on benchmarks including MMLU, CMMLU, BBH, and GSM8k, and the Yi-VL-34B vision-language model ranked first among open-source models on MMMU and CMMMU based on data available up to January 2024.[9](https://github.com/01-ai/Yi/blob/main/README.md?plain=1)
**Long context.** After a March 2024 enhancement, the Yi-34B-200K's score on the "Needle-in-a-Haystack" retrieval test rose 10.5%, from 89.3% to 99.8%.[5](https://github.com/01-ai/yi)
**Releases.** The Yi-1.5 series was open-sourced on May 13, 2024, improving coding, math, reasoning, and instruction-following abilities; the Yi-9B-200K followed in March 2024, and the Yi Cookbook 1.0, with tutorials in Chinese and English, was released in July 2024.[5](https://github.com/01-ai/yi) The Yi-1.5-34B-Chat model was updated on [Hugging Face](https://www.edgepedia.org/hugging-face) in August 2024.[10](https://huggingface.co/01-ai)
## Yi-Lightning
In October 2024, 01.AI released Yi-[Lightning](https://www.edgepedia.org/lightning), a 100B-parameter Mixture-of-Experts model that ranked sixth globally and first in China on the LMSYS blind test, which the company describes as a new historical best for China.[1](https://www.01.ai/) The December 2024 technical report states that Yi-Lightning ranked 6th overall on Chatbot Arena with 2nd to 4th place results in specialized categories including Chinese, Math, Coding, and Hard Prompts. It uses an enhanced MoE architecture with advanced expert segmentation and routing and optimized KV-caching, and was developed through pre-training, supervised fine-tuning, and reinforcement learning from human feedback, with a four-component safety framework called RAISE (Responsible AI Safety Engine). The report also notes a notable disparity between static benchmark results and real-world human preferences.[11](https://arxiv.org/abs/2412.01253)
## API platform and enterprise products
The 01.AI API platform lets developers use the Yi models for text generation, translation, summarization, logical reasoning, mathematical calculation, and code generation. It is highly compatible with the [OpenAI API](https://www.edgepedia.org/openai-api), so developers need minimal code changes to migrate.[12](https://platform.01.ai/docs)
The company's focus has shifted toward enterprise AI transformation. In March 2025 it released the WorldWise Enterprise LLM Platform (known in Chinese as Wanzhi, 万智), built on its self-developed end-to-end foundation-model technology stack and covering model deployment, application practice, and fine-tuning tools. In July 2026 it released TrueNorth, an enterprise AI decision hub platform extended into Boss AI, Investor AI, and TopSales AI. In January 2026 the company was recognized as a National High-Tech Enterprise and a Beijing Specialized, Refined, Distinctive, and Innovative SME.[1](https://www.01.ai/) The WorldWise platform, listed in Hong Kong's Smart Government Innovation LAB catalog by 01.AI (HK) Limited, embeds open-source models such as DeepSeek into enterprise ecosystems with data-security controls and supports on-premise deployment; a described use case is a customer-service assistant for a bank's wealth-management service in [Greater China](https://www.edgepedia.org/greater-china).[13](https://www.smartlab.gov.hk/en/ai_solutions/a-0107)
The company also describes sovereign AI offerings and industry solutions for supply chain and logistics, manufacturing, energy, agriculture, investment, education, and retail, along with a strategy-consultation and Forward Deployed Engineer co-creation model for enterprise transformation.[1](https://www.01.ai/)
## Context in Chinese AI
01.AI is counted among the group of Chinese AI unicorns, alongside StepFun, Zhipu AI, Moonshot, and MiniMax, that rose to prominence as Chinese frontier models gained global attention.[14](https://time.com/7358175/china-us-ai-race/) A Wall Street Journal report in late 2023 described the company as China's answer to OpenAI.[15](https://www.wsj.com/public/resources/documents/HIZ1FMj7pTV5OmEijmTg-WSJNewsPaper-11-30-2023.pdf)
## References
1. [01.AI official website, About Us and company timeline](https://www.01.ai/)
2. [Chinese AI startup 01.AI looks to raise $200 million, Reuters via Yahoo Tech, December 14, 2023](https://tech.yahoo.com/ai/articles/chinese-ai-startup-01-ai-092630274.html)
3. [01.AI API Platform, User Agreement](https://platform.01.ai/useragreement)
4. [Yi: Open Foundation Models by 01.AI, arXiv:2403.04652, March 2024](https://arxiv.org/abs/2403.04652)
5. [01-ai/Yi repository, GitHub](https://github.com/01-ai/yi)
6. [Chinese AI startup 01.AI looks to raise $200 mln, Reuters via Yahoo Finance, December 14, 2023](https://finance.yahoo.com/news/chinese-ai-startup-01-ai-092104117.html)
7. [Kai-Fu Lee, TIME, September 2023](https://qa.time.com/6310603/kai-fu-lee/)
8. [Valued at $1B, Kai-Fu Lee's LLM startup unveils open source model, TechCrunch, November 5, 2023](https://techcrunch.com/2023/11/05/valued-at-1b-kai-fu-lees-llm-startup-unveils-open-source-model/)
9. [Yi/README.md, GitHub](https://github.com/01-ai/Yi/blob/main/README.md?plain=1)
10. [01-ai organization, Hugging Face](https://huggingface.co/01-ai)
11. [Yi-Lightning Technical Report, arXiv:2412.01253, December 2024](https://arxiv.org/abs/2412.01253)
12. [01.AI API Platform documentation](https://platform.01.ai/docs)
13. [01.AI Worldwise Enterprise LLM Platform, Smart Government Innovation LAB, Hong Kong](https://www.smartlab.gov.hk/en/ai_solutions/a-0107)
14. [How China Caught Up on AI, TIME](https://time.com/7358175/china-us-ai-race/)
15. [Wall Street Journal, November 30, 2023, front-page PDF](https://www.wsj.com/public/resources/documents/HIZ1FMj7pTV5OmEijmTg-WSJNewsPaper-11-30-2023.pdf)
---
*Topic: Encyclopedia › Technology and the built world › Computing and digital systems › Modern AI: foundation models, generative AI, and the AI industry › AI companies, people, and products › Chinese AI companies*
*Initially written Sep 17, 2026 · Reviewed: Sep 18, 2026; Oct 15, 2026 · Edited: Sep 18, 2026; Oct 11, 2026; Oct 15, 2026 · Last review: Oct 15, 2026*
*Copyright 2026 EdgeChat AI, a subsidiary of Biostate AI.*
License: Edgepedia Community License 1.0, https://www.edgepedia.org/edgepedia/license