LLM EXPLORER 59,516 MODELS INDEXED

Suzume Llama 3 8B Multilingual by lightblue

By lightblue · 8215 downloads
HF Repo ↗ Copy URL

Suzume Llama 3 8B Multilingual is an open-source language model by lightblue. Features: 8b LLM, VRAM: 16.1GB, Context: 8K, License: other, Instruction-Based, LLM Explorer Score: 0.24, Arc: 60.8, HellaSwag: 79.5, MMLU: 66.6, GSM8K: 61.6.

  Arxiv:2405.12612 Base model:finetune:meta-llama... Base model:meta-llama/meta-lla...   Conversational   Deploy:azure   Endpoints compatible   Generated from trainer   Instruct   Llama   Pytorch   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/lightblue/suzume-llama-3-8B-multilingual 

Suzume Llama 3 8B Multilingual Parameters and Internals

Model Type 
multilingual, text generation
Use Cases 
Areas:
Research, Commercial applications
Applications:
Multilingual chat applications
Primary Use Cases:
Multilingual conversational AI
Limitations:
Excludes certain problem categories for Russian, Not yet fully evaluated
Considerations:
Ongoing evaluation and feedback encouraged
Additional Notes 
Ongoing development with future releases
Supported Languages 
languages_supported (multilingual), proficiency_levels (high)
Training Details 
Data Sources:
lightblue/tagengo-gpt4, lmsys/lmsys-chat-1m, megagonlabs/instruction_ja, openchat/openchat_sharegpt4_dataset
Data Volume:
90,000 multilingual conversations
Methodology:
finetuning
Context Length:
8192
Training Time:
2.5 hours
Hardware Used:
4 x A100 (80GB) GPUs
Model Architecture:
LlamaForCausalLM
Input Output 
Input Format:
Prompt messages should be constructed in a chat format
Accepted Modalities:
text
Output Format:
Text generation in response format
Performance Tips:
Utilize vLLM for optimal inference speed
LLM NameSuzume Llama 3 8B Multilingual
Repository 🤗https://huggingface.co/lightblue/suzume-llama-3-8B-multilingual 
Base Model(s)  Meta Llama 3 8B Instruct   meta-llama/Meta-Llama-3-8B-Instruct
Model Size8b
Required VRAM16.1 GB
Updated2026-08-07
Maintainerlightblue
Model Typellama
Instruction-BasedYes
Model Files  5.0 GB: 1-of-4   5.0 GB: 2-of-4   4.9 GB: 3-of-4   1.2 GB: 4-of-4   16.1 GB
Model ArchitectureLlamaForCausalLM
Licenseother
Context Length8192
Model Max Length8192
Transformers Version4.38.2
Tokenizer ClassPreTrainedTokenizerFast
Padding Token<|end_of_text|>
Vocabulary Size128256
Torch Data Typebfloat16

Quantized Models of the Suzume Llama 3 8B Multilingual

Model
Likes
Downloads
VRAM
Suzume Llama 3 8B Multilingual0174 GB

Best Alternatives to Suzume Llama 3 8B Multilingual

Best Alternatives
Context / RAM
Downloads
Likes
...otron 8B UltraLong 4M Instruct4192K / 32.1 GB1135125
UltraLong Thinking4192K / 16.1 GB23
...a 3.1 8B UltraLong 4M Instruct4192K / 32.1 GB17624
...a 3.1 8B UltraLong 2M Instruct2096K / 32.1 GB8759
...otron 8B UltraLong 2M Instruct2096K / 32.1 GB12418
Cthulhu 8B V1.41048K / 16.1 GB1010
...raLong 1M Instruct Abliterated1048K / 32.1 GB49
...a 3.1 8B UltraLong 1M Instruct1048K / 32.1 GB138729
...otron 8B UltraLong 1M Instruct1048K / 32.1 GB70259
Zero Llama 3.1 8B Beta61048K / 16.1 GB71