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

Dolphin 2.9.1 Mixtral 1x22b by cognitivecomputations

 »  All LLMs  »  cognitivecomputations  »  Dolphin 2.9.1 Mixtral 1x22b   URL Share it on

  6633   
  Autotrain compatible   Axolotl Base model:finetune:mistral-co... Base model:mistral-community/m...   Conversational Dataset:abacusai/systemchat-1.... Dataset:cognitivecomputations/... Dataset:cognitivecomputations/... Dataset:cognitivecomputations/...   Dataset:internlm/agent-flan Dataset:locutusque/function-ca... Dataset:m-a-p/codefeedback-fil... Dataset:microsoft/orca-math-wo...   Dataset:teknium/openhermes-2.5   En   Endpoints compatible   Generated from trainer   Mixtral   Moe   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/cognitivecomputations/dolphin-2.9.1-mixtral-1x22b 

Dolphin 2.9.1 Mixtral 1x22b Benchmarks

ARC: 58.53 vs 96.7 (so35)-39.5%
HellaSwag: 77.78 vs 95.3 (gpt4)-18.4%
MMLU: 59.14 vs 88.3 (so35)-33%
TruthfulQA: 54.75 vs 59 (gpt4)-7.2%
WinoGrande: 73.48 vs 87.5 (gpt4)-16%
GSM8K: 67.48 vs 96.4 (so35)-30%
LLME Score: 0.25186
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Dolphin 2.9.1 Mixtral 1x22b (cognitivecomputations/dolphin-2.9.1-mixtral-1x22b)

Dolphin 2.9.1 Mixtral 1x22b Parameters and Internals

Use Cases 
Primary Use Cases:
Instruction, conversational, and coding skills; initial agentic abilities; supports function calling
Limitations:
Highly compliant with any requests, even unethical ones, Uncensored, with filtered dataset for more compliance
Considerations:
Implement own alignment layer before deployment.
Additional Notes 
This model is based on Dolphin-2.9-Mixtral-8x22b and retains much of the original model's performance.
Supported Languages 
en (English)
Training Details 
Data Sources:
GPT-4 and other models
Methodology:
The model was fine-tuned, targeting all layers, and is an extracted expert using SLERP and a custom script. It uses an uncensored dataset with alignment and bias filtered out.
Context Length:
64000
Training Time:
27 hours
Hardware Used:
8xH100 GPUs provided by Crusoe Cloud
Model Architecture:
Mixtral architecture, not fully converted to dense model to retain performance.
LLM NameDolphin 2.9.1 Mixtral 1x22b
Repository 🤗https://huggingface.co/cognitivecomputations/dolphin-2.9.1-mixtral-1x22b 
Base Model(s)  mistral-community/Mixtral-8x22B-v0.1   mistral-community/Mixtral-8x22B-v0.1
Model Size22.2b
Required VRAM44.7 GB
Updated2025-02-22
Maintainercognitivecomputations
Model Typemixtral
Model Files  4.9 GB: 1-of-9   5.0 GB: 2-of-9   5.0 GB: 3-of-9   4.9 GB: 4-of-9   5.0 GB: 5-of-9   5.0 GB: 6-of-9   4.9 GB: 7-of-9   5.0 GB: 8-of-9   5.0 GB: 9-of-9
Supported Languagesen
Model ArchitectureMixtralForCausalLM
Licenseapache-2.0
Context Length65536
Model Max Length65536
Transformers Version4.40.2
Tokenizer ClassLlamaTokenizer
Padding Token</s>
Vocabulary Size32002
Torch Data Typebfloat16

Rank the Dolphin 2.9.1 Mixtral 1x22b 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? 43470 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 v20241227
  Support LLM Explorer