# wenge-research/YAYI2

YAYI 2 是中科闻歌研发的新一代开源大语言模型，采用了超过 2 万亿 Tokens 的高质量、多语言语料进行预训练。(Repo for YaYi 2 Chinese LLMs)

Repository: https://github.com/wenge-research/YAYI2
Canonical: https://ross.abutalabs.com/products/yayi2
Language: Python
License: Apache-2.0
License Family: permissive
Topics: artificial-intelligence, chat, chinese, gpt, natural-language-generation, pretrained-language-model, yayi
Archived: true
Last push: 2024-04-07T08:27:39+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 70
- inputs: {"age_days": 992, "days_push": 878, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2796, forks 18 (observed 2026-08-28T04:07:22.345881+00:00)

## What it is
YAYI 2 is a family of open-source multilingual large language models (30B Base and Chat variants) developed by Wenge Research, pretrained on over 2 trillion tokens with a focus on Chinese. The repository provides model weights, inference code, fine-tuning scripts (full-parameter and LoRA), and a 500GB pretraining dataset.

## Use cases
- run a chinese open-source llm locally
- download pretrained 30b language model weights
- finetune a chinese llm with lora
- chat with a chinese gpt-style model
- get multilingual llm pretraining data
- evaluate llm on c-eval and mmlu benchmarks

## When to choose
- you need a strong open-source Chinese-language LLM with permissive Apache-2.0 code licensing
- you want to fine-tune a 30B model with provided LoRA or full-parameter training scripts
- you need multilingual pretraining data for LLM research

## When to avoid
- you need a small model that runs on consumer hardware
- you require the Chat variant, which was not fully released at last update
- you need a commercially permissive model license (model weights use a custom YAYI license, data is CC BY-NC 4.0)

## Facets
- artifact type: learning-resource
- maturity: maintenance
- function: llm-inference, llm-training, chatbot
- domain: large-language-models, artificial-intelligence
- platform: python
- tags: chinese-llm, pretrained-language-model, 30b, multilingual, huggingface, model-weights, lora-finetuning, natural-language-processing, gpu, linux

## Member repositories
- wenge-research/YAYI2 (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:22.345881+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T08:14:39.956966+00:00, confidence not recorded.
  - readme: https://github.com/wenge-research/YAYI2 (fetched 2026-08-28T04:07:22.345881+00:00, sha d551f53b4a8f)
- Data as of 2026-08-30T08:39:29.467469+00:00.