| <div align="center"> | |
| <picture> | |
| <img alt="specify theme context for images" src="https://raw.githubusercontent.com/01-ai/Yi/main/assets/img/Yi_logo_icon_light.svg" width="150px"> | |
| </picture> | |
| </div> | |
| <p align="center"> | |
| <a href="https://huggingface.co/01-ai">🤗 HuggingFace</a> • | |
| <a href="https://www.modelscope.cn/organization/01ai/">🤖 ModelScope</a> • | |
| <a href="https://wisemodel.cn/organization/01.AI">✡️ WiseModel</a> | |
| <br/> | |
| <a href="https://github.com/01-ai/Yi-1.5">🐙 GitHub</a> • | |
| <a href="https://discord.gg/hYUwWddeAu">👾 Discord</a> • | |
| <a href="https://twitter.com/01ai_yi">🐤 Twitter</a> • | |
| <a href="https://github.com/01-ai/Yi/issues/43#issuecomment-1827285245">💬 WeChat</a> | |
| <br/> | |
| <a href="https://arxiv.org/abs/2403.04652">📝 Paper</a> • | |
| <a href="https://github.com/01-ai/Yi/tree/main?tab=readme-ov-file#faq">🙌 FAQ</a> • | |
| <a href="https://github.com/01-ai/Yi/tree/main?tab=readme-ov-file#learning-hub">📗 Learning Hub</a> | |
| </p> | |
| # Intro | |
| Yi-1.5 is an upgraded version of Yi. It is continuously pre-trained on Yi with a high-quality corpus of 500B tokens and fine-tuned on 3M diverse fine-tuning samples. | |
| Compared with Yi, Yi-1.5 delivers stronger performance in coding, math, reasoning, and instruction-following capability, while still maintaining excellent capabilities in language understanding, commonsense reasoning, and reading comprehension. | |
| <div align="center"> | |
| Model | Context Length | Pre-trained Tokens | |
| | :------------: | :------------: | :------------: | | |
| | Yi-1.5 | 4K | 3.6T | |
| </div> | |
| # Models | |
| - Chat models | |
| <div align="center"> | |
| Model | Download | |
| | :------------: | :------------: | | |
| Yi-1.5-34B-Chat | • 🤗 Hugging Face | |
| Yi-1.5-9B-Chat | • 🤗 Hugging Face | |
| Yi-1.5-6B-Chat | • 🤗 Hugging Face | |
| </div> | |
| - Base models | |
| <div align="center"> | |
| Model | Download | |
| | :------------: | :------------: | | |
| Yi-1.5-34B | • 🤗 Hugging Face | |
| Yi-1.5-9B | • 🤗 Hugging Face | |
| Yi-1.5-6B | • 🤗 Hugging Face | |
| </div> | |
| # Benchmarks | |
| - Chat models | |
| Waiting for benchmark results. | |
| - Base models | |
| - Yi-1.5-34B excels beyond or is on par with some larger models in overall performance. | |
| | Model | MMLU | CMMLU | BBH | AGIEval | HumanEva(+) | MBPP(+) | GSM8k | MATH | | |
| | --------------- | ---- | ----- | ---- | ------- | ----------- | ---------- | ----- | ---- | | |
| | Mistral 8\*22B | 77.8 | 60.4 | 61.9 | 58.6 | 45.1(34.1) | 71.2(-) | 81.7 | 41.8 | | |
| | DeepSeek-V2 | 78.5 | 84.0 | 78.9 | \- | 48.8(-) | 66.6(-) | 79.2 | 43.6 | | |
| | Qwen1.5-32B | 73.7 | 83 | 66.8 | 72.6 | 37.8(35.4) | 49.4(44.4) | 79.5 | 37.7 | | |
| | Qwen1.5-72B | 77.5 | 84.2 | 65.5 | 71.7 | 41.5(39.0) | 53.4(43.6) | 81.4 | 41.8 | | |
| | Llama3_70B_base | 78.7 | 68.9 | 65.0 | 52.8 | 38.4(34.8) | 69.7(52.9) | 82.4 | 42.4 | | |
| | Yi 1.5-34B | 77.1 | 84.8 | 76.4 | 71.1 | 46.3(40.2) | 65.5(55.4) | 82.7 | 41.0 | | |
| - Yi-1.5-9B is a strong performer among similarly sized open-source models. | |
| | Model | MMLU | CMMLU | BBH | AGIEval | HumanEval(+) | MBPP(+) | GSM8k | Math | | |
| | -------------- | ---- | ----- | ---- | ------- | ------------ | ---------- | ----- | ----- | | |
| | Gemma-7B | 64.3 | 48.4 | 41.1 | 46.0 | 33.5(28.0) | 45.8(32.8) | 55.7 | 24.8 | | |
| | Qwen1.5-7B | 61.0 | 73.4 | 33.4 | 61.6 | 36.0(31.1) | 46.1(37.6) | 70.1 | 20.3 | | |
| | Mistral-7B | 62.5 | 44.6 | 45.0 | 42.4 | 29.3(22.6) | 50.2(32.1) | 47.5 | 15.5 | | |
| | Mistral 8\*7B | 70.6 | 53.0 | 52.4 | 49.5 | 40.2(31.1) | 60.7(31.1) | 65.7 | 28.4 | | |
| | Llama3-8B_Base | 66.6 | 50.9 | 47.9 | 44.7 | 34.7(31.7) | 48.0(44.9) | 54.7 | 21.16 | | |
| | Yi 1.5-6B | 63.5 | 70.8 | 45.7 | 56.0 | 36.5(28.7) | 56.8(46.9) | 62.2 | 28.42 | | |
| | Yi 1.5-9B | 69.5 | 74.8 | 50.9 | 62.7 | 41.4(34.1) | 61.1(53.6) | 73.7 | 32.6 | | |
| # Quick Start | |
| For getting up and running with Yi-1.5 models quickly, see [README](https://github.com/01-ai/Yi-1.5). | |