Views
No views yet
| Architecture | Mixture-of-Experts (MoE) |
| Total Parameters | 3x7B |
| Activated Parameters | 3B |
| Experts Shared Frequency | 3 |
| Number of Layers (Dense layer included) | 31 |
| Number of Dense Layers | 1 |
| Attention Hidden Dimension | 2048 |
| MoE Hidden Dimension (per Expert) | 1408 |
| Number of Attention Heads | 16 |
| Number of Experts | 64 |
| Selected Experts per Token | 6 |
| Number of Shared Experts | 4 |
| Vocabulary Size | 128,880 |
| Context Length | 32K |
| Base Frequency of RoPE | 5,000,000 |
| Attention Mechanism | GQA |
| Activation Function | SwiGLU |
| Benchmark | Metric | Megrez2-3x7B -A3B-Preview | Qwen2.5-3B | Qwen2.5-7B | Qwen3-4B | Qwen3-8B | Phi-4-mini | Gemma-3-4B | GPT-4o-mini 2024-07-18 |
|---|---|---|---|---|---|---|---|---|---|
| Activate Params (B) | 3.0 | 3.1 | 7.6 | 4.0 | 8.2 | 3.8 | 4.3 | - | |
| Stored Params (B) | 7.5 | 3.1 | 7.6 | 4.0 | 8.2 | 3.8 | 4.3 | - | |
| General Tasks | |||||||||
| C-EVAL | EM | 91.7 | 68.2 | 76.2 | 72.2 | 77.9 | 40.0 | - | 66.3 |
| MMLU-Pro | EM | 67.6 | 43.7 | 56.3 | - | - | 52.8 | 43.6 | - |
| Instruction Tasks | |||||||||
| IF-Eval | Prompt Strict | 80.2 | 58.2 | 71.2 | 81.2 | 83.0 | 68.6 | 90.2 | 80.4 |
| Math & STEM Tasks | |||||||||
| MATH-500 | EM | 81.6 | 65.9 | 75.5 | 84.8 | 87.4 | 64.0 | 75.6 | 78.2 |
| GSM8K | EM | 83.6 | 86.7 | 91.6 | - | 93.2 | 88.6 | 89.2 | - |
| Coding Tasks | |||||||||
| HumanEval | Pass@1 | 74.4 | 74.4 | 84.8 | - | 85.9 | 74.4 | 71.3 | 87.2 |
| MBPP | Pass@1 | 88.0 | 72.7 | 79.2 | - | 77.0 | 65.3 | 63.2 | - |
transformers is recommended or transformers>=4.52.4 is required.
The following contains a code snippet illustrating how to use the model generate content based on given inputs.1from transformers import AutoModelForCausalLM, AutoTokenizer
2import torch
3
4path = "Infinigence/Megrez2-3x7B-A3B-Preview"
5device = "cuda"
6
7tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True)
8model = AutoModelForCausalLM.from_pretrained(path, torch_dtype=torch.bfloat16, device_map=device, trust_remote_code=True)
9
10messages = [
11 {"role": "user", "content": "世界上最高的山峰是哪座？"},
12]
13model_inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True).to(device)
14
15model_outputs = model.generate(
16 model_inputs,
17 do_sample=True,
18 max_new_tokens=1024
19)
20
21output_token_ids = [
22 model_outputs[i][len(model_inputs[i]):] for i in range(len(model_inputs))
23]
24
25responses = tokenizer.batch_decode(output_token_ids, skip_special_tokens=True)[0]
26print(responses)
27
28# 世界上最高的山峰是珠穆朗玛峰（Mount Everest），位于喜马拉雅山脉的中尼边境。珠穆朗玛峰的海拔高度为8,848.86米（29,031.7英尺），这一数据是由中国和尼泊尔在2020年共同宣布的最新测量结果。珠穆朗玛峰不仅是登山爱好者的圣地，也是地理和科学研究的重要对象。ModelScope adopts Python API similar to (though not entirely identical to) Transformers. For basic usage, simply modify the first line of the above code as follows:from modelscope import AutoModelForCausalLM, AutoTokenizervLLM and SGLang as inference backends. For more information, please visit the gitHub repository.1@misc{li2025megrez2technicalreport,
2 title={Megrez2 Technical Report},
3 author={Boxun Li and Yadong Li and Zhiyuan Li and Congyi Liu and Weilin Liu and Guowei Niu and Zheyue Tan and Haiyang Xu and Zhuyu Yao and Tao Yuan and Dong Zhou and Yueqing Zhuang and Bo Zhao and Guohao Dai and Yu Wang},
4 year={2025},
5 eprint={2507.17728},
6 archivePrefix={arXiv},
7 primaryClass={cs.CL},
8 url={https://arxiv.org/abs/2507.17728},
9}