Yi Model
01.AI's Yi model family — frontier development paused, company pivoted to enterprise
Overview
Yi is a large language model family developed by 01.AI, a Chinese AI startup founded by Kai-Fu Lee. Released in November 2023, the Yi-6B and Yi-34B models achieved top rankings on the HuggingFace Open LLM Leaderboard for their size class. Later releases Yi-Large and Yi-Lightning (both late 2024) remain 01.AI's most recent own-trained models.
In March 2025, 01.AI stopped pretraining its own large frontier models and pivoted to selling enterprise AI solutions built on DeepSeek's models via its "WorldWise Enterprise LLM Platform" (deployment and fine-tuning tooling), framing 2025 as a "foundation year" and 2026 as a "deployment era." The company today operates more as an AI systems-integrator than an active LLM-pretraining lab; the existing Yi weights remain available and self-hostable, but should not be expected to receive further frontier updates.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
Use cases
ASEAN Perspective
Yi Model in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Yi is the LLM family from 01.AI, offered as open weights with notably strong bilingual Chinese-English performance that made it one of the more respected open-model lineages of 2023–24. For teams needing solid Chinese-language capability in a self-hostable package, the existing Yi models remain credible. The essential context: 01.AI stopped pretraining new frontier models in March 2025 and now sells enterprise and sovereign-AI solutions built on DeepSeek's models rather than its own — a pivot that's winning business (a reported Rmb 1.5bn in 2026 contracts) but means Yi itself is frozen. Yi-Large and Yi-Lightning, from late 2024, are the last in-house models. Treat Yi as a mature, static family; for current Chinese open models, DeepSeek or Qwen is the better pick — 01.AI's own product line now agrees.
What people say
Kai-Fu Lee's math is blunt: pretraining a frontier model only pays if you can amass hundreds of millions of users, and 01.AI couldn't. So in March 2025 the company stopped pretraining altogether and rebuilt itself as an enterprise integrator on top of DeepSeek's open models, customising them for finance, gaming and legal clients — and, increasingly, chasing sovereign-AI deals with governments. The bet is working commercially: roughly Rmb 500m in 2025 orders, a reported Rmb 1.5bn in 2026 contracts, a fresh funding round in the works and a 2027 IPO on the roadmap.
That leaves the Yi models themselves in an odd spot. Yi-6B and Yi-34B topped the Hugging Face Open LLM Leaderboard for their size class at release in November 2023, and the bilingual Chinese-English quality was genuinely strong — it's why the family earned its reputation. Yi-Large and Yi-Lightning, both late 2024, are the last models 01.AI trained itself. The weights are still downloadable and self-hostable, and Yi 1.5 variants still turn up in production where teams standardised early and saw no reason to migrate.
But "still works" is not "still advancing." No further pretraining means no Yi answer to DeepSeek-V3-class or Qwen3-class releases — 01.AI now ships DeepSeek to its own customers. Use the existing Yi weights if you already run them or need a known-quantity bilingual model. Starting fresh in 2026, DeepSeek or Qwen is the obvious pick, a conclusion 01.AI has effectively endorsed with its own product line.
Summary of public user & expert reviews, compiled by RECATOOLS.
Notable facts
- 01.AI raised $200 million at a $1 billion valuation in 6 months — one of the fastest unicorn valuations in Chinese AI history.
- Yi-34B was trained on 3 trillion tokens, significantly more than comparable Western models at its release, giving it broader knowledge coverage.
- Kai-Fu Lee, Yi's founder, is the author of 'AI Superpowers', the most widely read book on US-China AI competition.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including Yi Model's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Yi Model unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Yi Model directly →
Spotted something out of date? Suggest an update →
Yi Model in the news
Alternatives to Yi Model
More in LLMs & Chat
