Awesome
🏰 LLM Zoo
As new animal species are being discovered in the world of natural language processing (NLP) 🌍 every day, it becomes necessary to establish a zoo 🦁 to accommodate them.
This project collects below information of various open- and closed-source LLMs (after the release of ChatGPT):
- Release time
- Model size
- Languages supported
- Domain
- Training data
- Links to resources: GitHub, HuggingFace, Demo, Paper, Official blog
📰 News
- [2023.05.03] First release! We will regularly update 🔄 the repository to keep track of the latest LLMs. We welcome 👐 any contributions to this project. Please feel free to open an issue or submit a pull request to include new LLMs or update the information of existing LLMs 🙏.
📖 Open-Sourced LLMs
Release Time | Model | Version | Size | Backbone | Langs | Domain | Training Data | GitHub | HF | Paper | Demo | Official Blog |
---|---|---|---|---|---|---|---|---|---|---|---|---|
2023.02.27 | LLaMA | llama-7b/13b/33b/65b | 7B/13B/33B/65B | - | en | General | <details><summary><b>detail</b></summary>1T tokens (English CommonCrawl, C4, Github, Wikipedia, Gutenberg and Books3, ArXiv, Stack Exchange)</details> | [link] | [link] | [link] | - | [link] |
2023.03.13 | Alpaca | alpaca-7b/13b | 7B/13B | LLaMA | en | General | <details><summary><b>detail</b></summary>52k instruction-following data generated by InstructGPT [link]</details> | [link] | [link] | - | [link] | [link] |
2023.03.13 | Vicuna | vicuna-7b/13b-delta-v1.1 | 7B/13B | LLaMA | en | General | <details><summary><b>detail</b></summary>70K samples from sharedGPT</details> | [link] | [link] | - | [link] | [link] |
2023.03.14 | ChatGLM | chatglm-6b | 6B | GLM | zh, en | General | <details><summary><b>detail</b></summary>supervised fine-tuning, feedback bootstrap, and reinforcement learning with human feedback</details> | [link] | [link] | - | - | [link] |
2023.03.14 | ChatGLM | chatglm-130b | 130B | GLM | zh, en | General | <details><summary><b>detail</b></summary>supervised fine-tuning, feedback bootstrap, and reinforcement learning with human feedback</details> | [link] | - | [link] | [link] | [link] |
2023.03.16 | Guanaco | - | 7B | LLaMA | ja, zh, en, de | General | <details><summary><b>detail</b></summary>multilingual datasets [link]</details> | [link] | [link] | - | - | - |
2023.03.24 | Dolly | dolly-v1-6b | 6B | GPT-J-6B | en | General | <details><summary><b>detail</b></summary>52k stanford alpaca instruction-following data [link]</details> | - | [link] | - | - | [link] |
2023.03.24 | ChatDoctor | - | 7B | LLaMA | en | Medicine | <details><summary><b>detail</b></summary>52K stanford alpaca [link], 100K HealthCareMagic [link], 10K icliniq [link], 5K GenMedGPT-5k [link]</details> | [link] | - | [link] | [link] | - |
2023.03.25 | LuoTuo | Chinese-alpaca-lora | 7B | LLaMA | zh, en | General | <details><summary><b>detail</b></summary>Translated 52k stanford alpaca instruction-following data [link], guanaco [link]</details> | [link] | [link] | - | - | - |
2023.03.26 | BELLE | BELLE-7B-0.2M/0.6M/1M/2M | 7B | BLOOMZ-7B1-mt | zh, en | General | <details><summary><b>detail</b></summary>0.2M/0.6M/1M/2M Chinese data [link], 52k stanford alpaca instruction-following data [link]</details> | [link] | [link] | [link] | - | - |
2023.03.28 | Linly (伶荔) | Linly-Chinese-LLaMA 7b/13b/33b | 7B/13B/33B | LLaMA | zh | General | <details><summary><b>detail</b></summary>Chinese-English parallel corpora [link], Chinese Wikipedia, community interaction, news data [link], scientific literature [link]</details> | [link] | [link] | - | - | - |
2023.03.28 | Linly (伶荔) | Linly-ChatFlow 7b/13b | 7B/13B | LLaMA | zh | General | <details><summary><b>detail</b></summary>BELLE [link], pCLUE [link], CSL [link], GuanacoDataset [link], Chain-of-Thought [link], news_commentary [link], firefly [link]</details> | [link] | [link] | - | - | [link] |
2023.04.01 | BAIZE | baize-7B/13B/30B | 7B/13B/30B | LLaMA | en | General | <details><summary><b>detail</b></summary>52K Stanford Alpaca [link], 54K Quora [link], 57K StackOverFlow [link]</details> | [link] | [link] | [link] | [link] | - |
2023.04.03 | Koala | - | 13B | LLaMA | en | General | <details><summary><b>detail</b></summary>ShareGPT, HC3 [link], OIG [link], Stanford alpaca [link], Anthropic HH [link], OpenAI WebGPT [link], OpenAI Summarization [link]</details> | - | [link] | - | [link] | [link] |
2023.04.03 | BAIZE | baize-healthcare-7b | 7B | LLaMA | en | Medicine | <details><summary><b>detail</b></summary>54K Quora [link], 47K medical dialogs [link]</details> | [link] | [link] | - | - | - |
2023.04.06 | Firefly (流萤) | firefly-1b4/2b6 | 1.4B/2.6B | BLOOM-ZH | zh | General | <details><summary><b>detail</b></summary>Chinese question-answering pairs [link], [link]</details> | [link] | [link] | - | - | - |
2023.04.08 | Phoenix | Phoenix-chat-7b | 7B | BLOOMZ | multi | General | <details><summary><b>detail</b></summary>conversation data [link]</details> | [link] | [link] | - | - | - |
2023.04.09 | Phoenix | Phoenix-inst-chat-7b | 7B | BLOOMZ | multi | General | <details><summary><b>detail</b></summary>conversation data [link], instruction data</details> | [link] | [link] | - | - | - |
2023.04.10 | Chimera | chimera-chat-7b/13b | 7B/13B | LLaMA | latin | General | <details><summary><b>detail</b></summary>conversation data [link]</details> | [link] | [link] | - | - | - |
2023.04.11 | Chimera | chimera-inst-chat-7b/13b | 7B/13B | LLaMA | latin | General | <details><summary><b>detail</b></summary>conversation data [link], instruction data</details> | [link] | [link] | - | - | - |
2023.04.12 | Dolly | dolly-v2-12b | 12B | pythia-12b | en | General | <details><summary><b>detail</b></summary>15k human-generated prompt/response pairs [link]</details> | [link] | [link] | - | - | [link] |
2023.04.14 | MedAlpaca | medalpaca 7b/13b | 7B/13B | LLaMA | en | Medicine | <details><summary><b>detail</b></summary>question-answering pairs from flash card, wikidoc, stackexchange and ChatDoctor</details> | [link] | [link] | [link] | - | - |
2023.04.19 | BELLE | BELLE-LLaMA-7B/13B-2M | 7B/13B | LLaMA | zh, en | General | <details><summary><b>detail</b></summary>2M Chinese data [link], 52k stanford alpaca instruction-following data [link]</details> | [link] | [link] | [link] | - | - |
2023.04.21 | MOSS | moss-moon-003-base | 16B | CodeGen | zh, en | General | <details><summary><b>detail</b></summary>100B Chinese tokens and 20B English tokens</details> | [link] | [link] | - | [link] | [link] |
2023.04.21 | MOSS | moss-moon-003-sft | 16B | moss-moon-003-base | zh, en | General | <details><summary><b>detail</b></summary>1.1M multi-turn conversational data (generated from ChatGPT) [link]</details> | [link] | [link] | - | [link] | [link] |
2023.04.21 | MOSS | moss-moon-003-sft-plugin | 16B | moss-moon-003-base | zh, en | General | <details><summary><b>detail</b></summary>1.1M multi-turn conversational data [link], 300K plugin-augmented data (generated by InstructGPT) [link]</details> | [link] | [link] | - | [link] | [link] |
2023.04.22 | HuggingChat | oasst-sft-6-llama-30b | 30B | LLaMA | multi | General | <details><summary><b>detail</b></summary>human-generated, human-annotated assistant-style conversation corpus consisting of 161k messages in 35 languages [link]</details> | [link] | [link] | - | [link] | - |
2023.06.19 | KnowLM | zhixi-13b | 13B | LLaMA | zh, en | General | <details><summary><b>detail</b></summary>human-generated, machine-generated and Knowledge Graph-generated in Chinese and English [link]</details> | [link] | [link] | - | - | - |
2023.06.21 | BayLing(百聆) | BayLing-7b/13b | 7B/13B | LLaMA | zh, en | General | <details><summary><b>detail</b></summary>160K human-generated, machine-generated multi-turn interactive translation corpus, alpaca instructions and sharegpt conversations [link]</details> | [link] | [link] | [link] | [link] | [link] |
2023.07.18 | LLaMA 2 | llama-2-7b/13b/70b-(chat) | 7B/13B/70B | - | en | General | <details><summary><b>detail</b></summary>2T tokens (Most in English, a new mix of data from publicly available sources)</details> | [link] | [link] | [link] | - | [link] |
📕 Closed-Sourced LLMs
Release Time | Model | Version | Size | Langs | Domain | Demo | Official Blog | Paper |
---|---|---|---|---|---|---|---|---|
2022.11.30 | ChatGPT | gpt-3.5-turbo | - | multi | general | [link] | [link] | - |
2023.03.14 | Claude | Claude Instant</br>Claude-v1 | - | multi | general | [link] | [link] | - |
2023.03.14 | GPT | gpt-4 | - | multi | general | [link] | [link] | [link] |
2023.03.16 | Ernie Bot (文心一言) | - | - | zh, en | general | [link] | [link] | - |
2023.03.21 | Bard | - | - | multi | general | [link] | [link] | - |
2023.03.30 | BloombergGPT | - | 50B | en | finance | - | [link] | [link] |
2023.04.11 | Tongyi Qianwen (通义千问) | - | - | multi | general | [link] | [link] | - |
2023.07.07 | OmModel(欧姆大模型) | - | - | multi | general | [link] | [link] | - |
2023.07.11 | Claude 2 | Claude-v2 | - | multi | general | - | [link] | [link] |
🏗 TODO List
- Include open-sourced LLMs
- Include closed-sourced LLMs
- Include a systematic review of common training data
- Include interesting use cases of various LLMs
- Performance of LLMs on various evaluation tasks
📝 Citation
If you find this repository useful, please consider citing.
@software{li2023llmzoo,
title = {LLM Zoo}
author = {Li, Xingxuan and Zhang, Wenxuan and Bing, Lidong},
url = {https://github.com/DAMO-NLP-SG/LLM-Zoo},
year = {2023}
}