LLM EXPLORER 59,358 MODELS INDEXED

StableBeluga2 by petals-team

By petals-team · 15937 downloads
HF Repo ↗ Copy URL

StableBeluga2 is an open-source language model by petals-team. Features: 69b LLM, VRAM: 1.7GB, Context: 4K, LLM Explorer Score: 0.12.

  Arxiv:2306.02707   Arxiv:2307.09288 Dataset:conceptofmind/cot subm... Dataset:conceptofmind/flan2021... Dataset:conceptofmind/niv2 sub... Dataset:conceptofmind/t0 submi...   En   Endpoints compatible   Llama   Region:us   Safetensors
Model Card on HF 🤗: https://huggingface.co/petals-team/StableBeluga2 

StableBeluga2 Parameters and Internals

Model Type 
auto-regressive language model, text-generation
Additional Notes 
The model uses `bfloat16` precision.
Supported Languages 
en (high)
Training Details 
Data Sources:
conceptofmind/cot_submix_original, conceptofmind/flan2021_submix_original, conceptofmind/t0_submix_original, conceptofmind/niv2_submix_original
Methodology:
Fine-tuning on Llama2 70B; supervised fine-tuning with Orca-style dataset
Input Output 
Input Format:
### System: System prompt ### User: Your prompt here ### Assistant:
Accepted Modalities:
text
Output Format:
Text generated based on input.
LLM NameStableBeluga2
Repository 🤗https://huggingface.co/petals-team/StableBeluga2 
Model Size69b
Required VRAM1.7 GB
Updated2026-07-31
Maintainerpetals-team
Model Typellama
Model Files  1.7 GB: 1-of-81   1.7 GB: 2-of-81   1.7 GB: 3-of-81   1.7 GB: 4-of-81   1.7 GB: 5-of-81   1.7 GB: 6-of-81   1.7 GB: 7-of-81   1.7 GB: 8-of-81   1.7 GB: 9-of-81   1.7 GB: 10-of-81   1.7 GB: 11-of-81   1.7 GB: 12-of-81   1.7 GB: 13-of-81   1.7 GB: 14-of-81   1.7 GB: 15-of-81   1.7 GB: 16-of-81   1.7 GB: 17-of-81   1.7 GB: 18-of-81   1.7 GB: 19-of-81   1.7 GB: 20-of-81   1.7 GB: 21-of-81   1.7 GB: 22-of-81   1.7 GB: 23-of-81   1.7 GB: 24-of-81   1.7 GB: 25-of-81   1.7 GB: 26-of-81   1.7 GB: 27-of-81   1.7 GB: 28-of-81   1.7 GB: 29-of-81   1.7 GB: 30-of-81   1.7 GB: 31-of-81   1.7 GB: 32-of-81   1.7 GB: 33-of-81   1.7 GB: 34-of-81   1.7 GB: 35-of-81   1.7 GB: 36-of-81   1.7 GB: 37-of-81   1.7 GB: 38-of-81   1.7 GB: 39-of-81   1.7 GB: 40-of-81   1.7 GB: 41-of-81   1.7 GB: 42-of-81   1.7 GB: 43-of-81
Supported Languagesen
Model ArchitectureLlamaForCausalLM
Context Length4096
Model Max Length4096
Transformers Version4.32.0.dev0
Vocabulary Size32000
Torch Data Typebfloat16

Best Alternatives to StableBeluga2

Best Alternatives
Context / RAM
Downloads
Likes
...ine Tu Ophtho 3 1e 05 Stanford4K / 138.7 GB50
V Alpha Tross4K / 138.7 GB7420
Uni TianYan V14K / 138 GB660
SaiLy Experiment V14K / 276 GB152
Airoboros L2 C70b 3.1.24K / 138 GB790
Model 007 Preview4K / 138 GB291
QCammel 70 X AWQ4K / 36.6 GB61
QCammel 70 X GPTQ4K / 35.3 GB102
Note: green Score (e.g. "73.2") means that the model is better than petals-team/StableBeluga2.