目前比較主流的 AI 模型

｜超大模型｜（硬體要求很高）

Databricks

databricks/dbrx-instruct · Hugging Facehuggingface

---

｜大模型｜（硬體要求稍高）

---

『Mistral AI』

mistralai/Mixtral-8x22B-Instruct-v0.1 · Hugging Facehuggingface

mistralai/Mixtral-8x22B-v0.1 · Hugging Facehuggingface

mistralai/Mixtral-8x7B-Instruct-v0.1 · Hugging Facehuggingface

Mixtral-8x7B-Instruct

mistralai/Mistral-7B-Instruct-v0.2 · Hugging Facehuggingface

Mistral-7B-Instruct-v0.2

mistralai/Mistral-7B-v0.1 · Hugging Facehuggingface

Mistral-7B-v0.1

---

『Google』

Gemma release - a google Collectionhuggingface

CodeGemma Release - a google Collectionhuggingface

---

『Hugging Face H4』

Zephyr 7B - a HuggingFaceH4 Collectionhuggingface

HuggingFaceH4/zephyr-7b-beta · Hugging Facehuggingface

zephyr-7b

tiiuae/falcon-7b · Hugging Facehuggingface

falcon-7b

『Meta』

meta-llama/Llama-2-7b · Hugging Facehuggingface

Llama-2-7b

---

[中國]

天工 Skywork

Skywork/Skywork-13B-base · Hugging Facehuggingface

中國清華

THUDM/cogvlm-chat-hf · Hugging Facehuggingface

（視覺）

THUDM/chatglm3-6b · Hugging Facehuggingface

｜小模型｜（硬體要求略低）

『Microsoft』

優點：資料精挑細選，模組小，佔用記憶體相對少，對硬體要求略低，易於部署。

Phi-3 - a microsoft Collectionhuggingface

microsoft/phi-2 · Hugging Facehuggingface

---

『簡體中文』

China models - a Ji-Xiang Collectionhuggingface

---

『繁體中文』

聯發科 （MediaTek）（MediaTek）

MediaTek-Research/Breeze-7B-Instruct-v0_1 · Hugging Facehuggingface

MediaTek-Research/Breeze-7B-Base-v0_1 · Hugging Facehuggingface

TAIDE

taide/TAIDE-LX-7B-Chat · Hugging Facehuggingface

taide/Llama3-TAIDE-LX-8B-Chat-Alpha1 · Hugging Facehuggingface

taide/Llama3-TAIDE-LX-8B-Chat-Alpha1-4bit · Hugging Facehuggingface

數位發展部

yentinglin/Taiwan-LLM-13B-v2.0-chat · Hugging Facehuggingface

yentinglin/Taiwan-LLM-8x7B-DPO · Hugging Facehuggingface