LLM EXPLORER
 LLM Explorer: A Curated Large Language Model Directory and Analytics
Dark Theme
  LLM  List    LLM Hosting    LLM Leaderboards    Blog    Newsfeed    Advertise

Smaug 34B V0.1 by abacusai

 »  All LLMs  »  abacusai  »  Smaug 34B V0.1   URL Share it on

  3718   
  Merged Model   Arxiv:2402.13228   Autotrain compatible Base model:finetune:jondurbin/... Base model:jondurbin/bagel-34b...   Conversational   Endpoints compatible   Llama   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/abacusai/Smaug-34B-v0.1 

Smaug 34B V0.1 Benchmarks

MMLU Pro: 39.37
GPQA: 10.63
MUSR: 8.13
BBH: 34.26
IFEval: 50.16 vs 88 (so35)-43%
ARC: 74.23 vs 96.7 (so35)-23.2%
HellaSwag: 86.76 vs 95.3 (gpt4)-9%
MMLU: 76.66 vs 88.3 (so35)-13.2%
TruthfulQA: 70.22 vs 59 (gpt4)19%
WinoGrande: 83.66 vs 87.5 (gpt4)-4.4%
GSM8K: 72.18 vs 96.4 (so35)-25.1%
LLME Score: 0.28779
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Smaug 34B V0.1 (abacusai/Smaug-34B-v0.1)

Smaug 34B V0.1 Parameters and Internals

Model Type 
finetune
Additional Notes 
Smaug-34B-v0.1 leverages a new fine-tuning technique, DPO-Positive (DPOP), which is superior in reducing the likelihood of the preferred examples being minimized while enhancing probability between preferred and dispreferred classes.
Training Details 
Data Sources:
ARC, HellaSwag, MetaMath
Methodology:
Fine-tuning using DPO-Positive (DPOP) and pairwise preference versions
LLM NameSmaug 34B V0.1
Repository 🤗https://huggingface.co/abacusai/Smaug-34B-v0.1 
Base Model(s)  jondurbin/bagel-34b-v0.2   jondurbin/bagel-34b-v0.2
Merged ModelYes
Model Size34b
Required VRAM69.2 GB
Updated2024-12-21
Maintainerabacusai
Model Typellama
Model Files  4.8 GB: 1-of-15   4.8 GB: 2-of-15   5.0 GB: 3-of-15   4.8 GB: 4-of-15   4.8 GB: 5-of-15   5.0 GB: 6-of-15   4.8 GB: 7-of-15   4.8 GB: 8-of-15   5.0 GB: 9-of-15   4.8 GB: 10-of-15   4.8 GB: 11-of-15   5.0 GB: 12-of-15   4.8 GB: 13-of-15   4.8 GB: 14-of-15   1.2 GB: 15-of-15
Model ArchitectureLlamaForCausalLM
Licenseapache-2.0
Context Length200000
Model Max Length200000
Transformers Version4.36.2
Tokenizer ClassLlamaTokenizer
Padding Token<unk>
Vocabulary Size64000
Torch Data Typebfloat16

Best Alternatives to Smaug 34B V0.1

Best Alternatives
Context / RAM
Downloads
Likes
Casual Magnum 34B195K / 68.8 GB131
34B Beta195K / 69.2 GB354862
Yi 34B 200K195K / 68.9 GB4525317
Bagel Hermes 34B Slerp195K / 68.9 GB39861
Bagel 34B V0.2195K / 68.7 GB395439
Yi 34B 200K AEZAKMI V2195K / 69.2 GB121012
Smaug 34B V0.1 ExPO195K / 69.2 GB29660
Faro Yi 34B195K / 69.2 GB37496
Bagel DPO 34B V0.5195K / 68.7 GB300817
Mergekit Slerp Anaazls195K / 69.2 GB80
Note: green Score (e.g. "73.2") means that the model is better than abacusai/Smaug-34B-v0.1.

Rank the Smaug 34B V0.1 Capabilities

🆘 Have you tried this model? Rate its performance. This feedback would greatly assist ML community in identifying the most suitable model for their needs. Your contribution really does make a difference! 🌟

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  

What open-source LLMs or SLMs are you in search of? 40013 in total.

Email us: info@extractum.io. Our Privacy Policy  | Terms and Conditions  | Suggest an improvement.
Our Social Media →  
Original data from HuggingFace, OpenCompass and various public git repos.
Release v20241217
  Support LLM Explorer