What are the hardware requirements for Yayi2 30B Llama?

Yayi2 30B Llama requires approximately 121.2 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 Yayi2 30B Llama and how large is it?

Yayi2 30B Llama is developed by cognitivecomputations, a model with 30b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Yayi2 30B Llama perform on standard benchmarks?

Yayi2 30B Llama has the following published scores: MMLU 70.6. Compare against reference models on this page or on the LLM Explorer leaderboards.

Yayi2 30B Llama by cognitivecomputations — VRAM 121.2GB, 4K context

Name: Yayi2 30B Llama
Author: cognitivecomputations

Yayi2 30B Llama is an open-source language model by cognitivecomputations. Features: 30b LLM, VRAM: 121.2GB, Context: 4K, License: other, LLM Explorer Score: 0.14, Arc: 35.7, HellaSwag: 53.4, MMLU: 70.6, GSM8K: 18.9.

Arxiv:2307.09288 Autotrain compatible En Endpoints compatible Llama Model-index Region:us Safetensors Sharded Tensorflow Zh

Model Card on HF 🤗: https://huggingface.co/cognitivecomputations/yayi2-30b-llama

Yayi2 30B Llama Parameters and Internals

Model Type

text-generation

Use Cases

Areas:

research, commercial applications

Limitations:

Potential biases based on training data

Additional Notes

The model uses advanced techniques like RLHF for better alignment with human values.

Supported Languages

zh (multilingual), en (multilingual)

Training Details

Data Sources:

multilingual data

Data Volume:

2.65 trillion tokens

Methodology:

Transformer-based, supervised fine-tuning with millions of instructions and reinforcement learning from human feedback (RLHF)

Context Length:

4096

Model Architecture:

64 layers, 64 heads, hidden size 7168, vocabulary size 81920

LLM Name	Yayi2 30B Llama
Repository 🤗	https://huggingface.co/cognitivecomputations/yayi2-30b-llama
Model Size	30b
Required VRAM	121.2 GB
Updated	2025-07-13
Maintainer	cognitivecomputations
Model Type	llama
Model Files	4.6 GB: 1-of-27 4.6 GB: 2-of-27 4.5 GB: 3-of-27 4.6 GB: 4-of-27 4.5 GB: 5-of-27 4.6 GB: 6-of-27 4.5 GB: 7-of-27 4.6 GB: 8-of-27 4.5 GB: 9-of-27 4.6 GB: 10-of-27 4.5 GB: 11-of-27 4.6 GB: 12-of-27 4.5 GB: 13-of-27 4.6 GB: 14-of-27 4.5 GB: 15-of-27 4.6 GB: 16-of-27 4.5 GB: 17-of-27 4.6 GB: 18-of-27 4.5 GB: 19-of-27 4.6 GB: 20-of-27 4.5 GB: 21-of-27 4.6 GB: 22-of-27 4.5 GB: 23-of-27 4.6 GB: 24-of-27 4.5 GB: 25-of-27 4.6 GB: 26-of-27 2.8 GB: 27-of-27
Supported Languages	zh en
Model Architecture	LlamaForCausalLM
License	other
Context Length	4096
Model Max Length	4096
Transformers Version	4.36.2
Tokenizer Class	LlamaTokenizer
Beginning of Sentence Token	<s>
End of Sentence Token	</s>
Unk Token	<unk>
Vocabulary Size	81920
Torch Data Type	float32

Rank the Yayi2 30B Llama 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

Quantized Models of the Yayi2 30B Llama

Model	Likes	Downloads	VRAM
Yayi2 30B Llama GGUF	10	511	12 GB
Yayi2 30B Llama GPTQ	2	12	17 GB

Best Alternatives to Yayi2 30B Llama

Best Alternatives	Context / RAM	Downloads	Likes
TildeOpen 30B	64K / 60.9 GB	4576	165
TildeOpen 30B 64K	64K / 60.9 GB	702	3
Tildeopen 30B Mu Instruct	64K / 60.9 GB	41	6
Flash Llama 30M 20001	32K / 0.1 GB	134	0
Smaug Slerp 30B V0.1	32K / 60.4 GB	79	0
Tenebra 30B Alpha01	16K / 65 GB	50	16
Llama33b 16K	16K / 65.2 GB	1	1
Veyra2 Mango 30M Base	4K / 0.1 GB	0	5
... Tokens By Perplexity Bottom K	4K / 5.4 GB	8	0
...via Sample With Temperature2.0	4K / 5.4 GB	8	0

Note: green Score (e.g. "73.2") means that the model is better than cognitivecomputations/yayi2-30b-llama.