What are the hardware requirements for Nanbeige2 16B Chat?

Nanbeige2 16B Chat requires approximately 31.6 GB of VRAM and supports a context window of 4K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Nanbeige2 16B Chat and how large is it?

Nanbeige2 16B Chat is developed by duoqi, a model with 16b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Nanbeige2 16B Chat perform on standard benchmarks?

Nanbeige2 16B Chat has the following published scores: MMLU 66.63. Compare against reference models on this page or on the LLM Explorer leaderboards.

Nanbeige2 16B Chat by duoqi — VRAM 31.6GB, 4K context

Name: Nanbeige2 16B Chat
Rating: 4 (8 reviews)
Author: duoqi

Nanbeige2 16B Chat is an open-source language model by duoqi. Features: 16b LLM, VRAM: 31.6GB, Context: 4K, License: apache-2.0, LLM Explorer Score: 0.17, Arc: 63.4, HellaSwag: 80.7, MMLU: 66.6, GSM8K: 63.9.

Autotrain compatible Conversational En Endpoints compatible Llama Region:us Safetensors Sharded Tensorflow Zh

Model Card on HF 🤗: https://huggingface.co/duoqi/Nanbeige2-16B-Chat

Nanbeige2 16B Chat Parameters and Internals

Model Type

text generation

Additional Notes

The model underwent initial training with 1 million samples through Supervised Fine-Tuning and subsequent curriculum learning on 400,000 challenging samples, with human feedback integrated through Direct Preference Optimization (DPO).

Supported Languages

en (fluent), zh (fluent)

Training Details

Data Sources:

4.5T tokens of high-quality training data

Data Volume:

4.5 trillion tokens

Methodology:

Supervised Fine-Tuning (SFT), Curriculum Learning, Direct Preference Optimization (DPO)

LLM Name	Nanbeige2 16B Chat
Repository 🤗	https://huggingface.co/duoqi/Nanbeige2-16B-Chat
Model Size	16b
Required VRAM	31.6 GB
Updated	2025-07-17
Maintainer	duoqi
Model Type	llama
Model Files	4.9 GB: 1-of-7 4.9 GB: 2-of-7 4.9 GB: 3-of-7 5.0 GB: 4-of-7 5.0 GB: 5-of-7 4.9 GB: 6-of-7 2.0 GB: 7-of-7
Supported Languages	en zh
Model Architecture	LlamaForCausalLM
License	apache-2.0
Context Length	4096
Model Max Length	4096
Transformers Version	4.28.1
Tokenizer Class	LlamaTokenizer
Padding Token	<unk>
Vocabulary Size	59392
Torch Data Type	bfloat16

Rank the Nanbeige2 16B Chat capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to Nanbeige2 16B Chat

Best Alternatives	Context / RAM	Downloads	Likes
...r SMB R1 Distill Llama 3.1 16B	128K / 32 GB	6	1
Phi 3.5 Mini Investigator 16B	128K / 7.6 GB	7	0
...aid Blackroot Grand HORROR 16B	8K / 33.8 GB	14	0
...aid Blackroot Grand HORROR 16B	8K / 33.8 GB	30	0
...o Sft Ties Post Merge Auto DPO	8K / 141.2 GB	15	0
Nanbeige 16B Base Llama	4K / 31.6 GB	9946	4
...ALAXY V03 Slimorca 1 Epoch 50k	4K / 31.8 GB	80	0
...ca 1 Epoch 50k DPO 1 Epoch 30k	4K / 31.8 GB	78	0
GALAXY XB V.03	4K / 31.9 GB	81	0
Llama 2 16B Nastychat	4K / 32.4 GB	1779	11

Note: green Score (e.g. "73.2") means that the model is better than duoqi/Nanbeige2-16B-Chat.