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

Pythia 6.9B HC3 by pszemraj

 »  All LLMs  »  pszemraj  »  Pythia 6.9B HC3   URL Share it on

  8   

Pythia 6.9B HC3 is an open-source language model by pszemraj. Features: 6.9b LLM, VRAM: 27.7GB, Context: 2K, License: apache-2.0, HF Score: 38.5, LLM Explorer Score: 0.1, Arc: 36.5, HellaSwag: 61.8, MMLU: 26.9, TruthfulQA: 45.1, WinoGrande: 60.8.

  Assistant Base model:eleutherai/pythia-6... Base model:finetune:eleutherai...   Chatgpt Dataset:pszemraj/hc3-textgen-q...   Generated from trainer   Gpt neox   Hc3   Pytorch   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/pszemraj/pythia-6.9b-HC3 

Pythia 6.9B HC3 Benchmarks

ARC: 36.52 vs 96.7 (so35)-62.2%
HellaSwag: 61.76 vs 95.3 (gpt4)-35.2%
MMLU: 26.94 vs 88.3 (so35)-69.5%
TruthfulQA: 45.05 vs 59 (gpt4)-23.6%
WinoGrande: 60.77 vs 87.5 (gpt4)-30.5%
LLME Score: 0.0978
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Pythia 6.9B HC3 (pszemraj/pythia-6.9b-HC3)
🌟 Advertise your project 🚀

Pythia 6.9B HC3 Parameters and Internals

Model Type 
text generation
Use Cases 
Areas:
research
Applications:
exploration into comparing RLHF tuning vs. 'guided'/specific tuning on 'quality' datasets
Limitations:
Not trained/fine-tuned with RLHF., Will not be as helpful/generalizable/safe as chatGPT., Model is ~30x smaller than chatGPT.
Additional Notes 
The model was not trained with Reinforcement Learning through Human Feedback (RLHF) and focuses on generating high-quality answers akin to what humans would desire.
Training Details 
Data Sources:
pszemraj/HC3-textgen-qa
Methodology:
fine-tuning
Input Output 
Accepted Modalities:
text
Output Format:
text
Performance Tips:
Contrastive search with top_k=4 and penalty_alpha=0.6 is used for generation.
Release Notes 
Version:
pythia-6.9b-hc3-qa-assistant
Notes:
Tested on causal language modeling with a training loss of 1.2598, and validated with a loss of 1.2372 and an accuracy of 0.6769.
LLM NamePythia 6.9B HC3
Repository 🤗https://huggingface.co/pszemraj/pythia-6.9b-HC3 
Base Model(s)  EleutherAI/pythia-6.9b-deduped   EleutherAI/pythia-6.9b-deduped
Model Size6.9b
Required VRAM27.7 GB
Updated2026-04-09
Maintainerpszemraj
Model Typegpt_neox
Model Files  3.8 GB: 1-of-8   3.8 GB: 2-of-8   4.0 GB: 3-of-8   3.9 GB: 4-of-8   3.8 GB: 5-of-8   3.8 GB: 6-of-8   3.8 GB: 7-of-8   0.8 GB: 8-of-8   3.8 GB: 1-of-8   3.8 GB: 2-of-8   4.0 GB: 3-of-8   3.9 GB: 4-of-8   3.8 GB: 5-of-8   3.8 GB: 6-of-8   3.8 GB: 7-of-8   0.8 GB: 8-of-8
Model ArchitectureGPTNeoXForCausalLM
Licenseapache-2.0
Context Length2048
Model Max Length2048
Transformers Version4.27.0.dev0
Tokenizer ClassGPTNeoXTokenizer
Vocabulary Size50432
Torch Data Typefloat32

Best Alternatives to Pythia 6.9B HC3

Best Alternatives
Context / RAM
Downloads
Likes
Pythia 6.9B Deduped 8K8K / 13.8 GB61
...I Pythia 6.9B Deduped Sft Tldr2K / 27.4 GB4740
Pythia 6.9B2K / 13.8 GB6960959
Ppo Tldr 6.9B2K / 13.8 GB80
Pythia 6.9B Deduped2K / 13.8 GB68738
...I Pythia 6.9B Deduped Sft Tldr2K / 13.8 GB400
Oasst Pythia 6.9B 4000 Steps2K / 13.8 GB8790
Open Instruct Pythia 6.9B Tulu2K / 27.6 GB7236
...an Large Pythia 6.9B Dev Phase2K / 27.6 GB33
....9B Deduped Synthetic Instruct2K / 27.5 GB26704
Note: green Score (e.g. "73.2") means that the model is better than pszemraj/pythia-6.9b-HC3.

Rank the Pythia 6.9B HC3 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? 52721 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.
Check out Ag3ntum — our secure, self-hosted AI agent for server management.
Release v20260328a
  Support LLM Explorer