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

Yi 34B 200K AEZAKMI RAW 2901 4bpw H6 EXL2 by waldie

 »  All LLMs  »  waldie  »  Yi 34B 200K AEZAKMI RAW 2901 4bpw H6 EXL2   URL Share it on

  3   
  Dataset:adamo1139/aezakmi v2   Dataset:adamo1139/rawrr v1   Endpoints compatible   Exl2   Llama   Quantized   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/waldie/Yi-34B-200K-AEZAKMI-RAW-2901-4bpw-h6-exl2 

Yi 34B 200K AEZAKMI RAW 2901 4bpw H6 EXL2 Benchmarks

LLME Score: 0.13315
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Yi 34B 200K AEZAKMI RAW 2901 4bpw H6 EXL2 (waldie/Yi-34B-200K-AEZAKMI-RAW-2901-4bpw-h6-exl2)
🌟 Advertise your project 🚀

Yi 34B 200K AEZAKMI RAW 2901 4bpw H6 EXL2 Parameters and Internals

Model Type 
chat
Use Cases 
Areas:
Chat
Limitations:
Not good at math, riddles, or complex reasoning
Considerations:
Repetition penalty recommended at 1.05, Temperature 1.0-1.2 advised, strong anti-refusal bias present.
Additional Notes 
Experimental model, has not been extensively tested. Uses "unsloth" for both DPO and SFT stages. Not overfitted, respects different system messages.
Training Details 
Data Sources:
RAWrr v1, AEZAKMI v2
Methodology:
Fine-tuned using DPO and SFT
Training Time:
DPO: 6 hours, SFT: 25 hours
Model Architecture:
Yi-34B-Llama, max_position_embeddings set at 4096 and then reverted back to 200K after applying LoRA.
Input Output 
Input Format:
ChatML format
Output Format:
Refusal-free responses
Performance Tips:
Set repetition penalty to 1.05 to avoid repetition.
LLM NameYi 34B 200K AEZAKMI RAW 2901 4bpw H6 EXL2
Repository 🤗https://huggingface.co/waldie/Yi-34B-200K-AEZAKMI-RAW-2901-4bpw-h6-exl2 
Model Size34b
Required VRAM18 GB
Updated2026-01-22
Maintainerwaldie
Model Typellama
Model Files  8.6 GB: 1-of-3   8.5 GB: 2-of-3   0.9 GB: 3-of-3
Quantization Typeexl2
Model ArchitectureLlamaForCausalLM
Licenseapache-2.0
Context Length200000
Model Max Length200000
Transformers Version4.36.2
Tokenizer ClassLlamaTokenizer
Padding Token<unk>
Vocabulary Size64000
Torch Data Typefloat16

Best Alternatives to Yi 34B 200K AEZAKMI RAW 2901 4bpw H6 EXL2

Best Alternatives
Context / RAM
Downloads
Likes
...34B 200K Aezakmi Raw 1902 EXL2195K / 20.7 GB61
Yi 34B 200K MAR2024 EXL2 4bpw195K / 18 GB41
Yi 34B 200K RPMerge195K / 68.9 GB5260
...B 200K RPMerge 4.65bpw H6 EXL2195K / 10.5 GB92
...mi Raw Toxic 2702 4.65bpw EXL2195K / 20.8 GB12
Opus V1 34B 5.0bpw H6 EXL2195K / 22.3 GB52
Opus V1 34B 8.0bpw H8 EXL2195K / 34.8 GB81
Opus V1 34B 4.0bpw H6 EXL2195K / 18 GB71
34B Beta 4.0bpw H6 EXL2195K / 18.1 GB81
34B Beta 5.0bpw H6 EXL2195K / 22.3 GB81
Note: green Score (e.g. "73.2") means that the model is better than waldie/Yi-34B-200K-AEZAKMI-RAW-2901-4bpw-h6-exl2.

Rank the Yi 34B 200K AEZAKMI RAW 2901 4bpw H6 EXL2 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? 51611 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 v20241124
  Support LLM Explorer