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

Facebook Opt 6.7B Qcqa Ub 16 Best For KV Cache by saarvajanik

 »  All LLMs  »  saarvajanik  »  Facebook Opt 6.7B Qcqa Ub 16 Best For KV Cache   URL Share it on

  255   

Facebook Opt 6.7B Qcqa Ub 16 Best For KV Cache is an open-source language model by saarvajanik. Features: 6.7b LLM, VRAM: 27.4GB, Context: 2K, License: mit, HF Score: 28.6, LLM Explorer Score: 0.17, Arc: 23.8, HellaSwag: 27.1, MMLU: 23.1, TruthfulQA: 46.7, WinoGrande: 50.8.

  En   Endpoints compatible   Opt   Pytorch   Region:us   Sharded
Model Card on HF 🤗: https://huggingface.co/saarvajanik/facebook-opt-6.7b-qcqa-ub-16-best-for-KV-cache 

Facebook Opt 6.7B Qcqa Ub 16 Best For KV Cache Benchmarks

ARC: 23.81 vs 96.7 (so35)-75.4%
HellaSwag: 27.05 vs 95.3 (gpt4)-71.6%
MMLU: 23.12 vs 88.3 (so35)-73.8%
TruthfulQA: 46.69 vs 59 (gpt4)-20.9%
WinoGrande: 50.83 vs 87.5 (gpt4)-41.9%
LLME Score: 0.16928
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Facebook Opt 6.7B Qcqa Ub 16 Best For KV Cache (saarvajanik/facebook-opt-6.7b-qcqa-ub-16-best-for-KV-cache)
🌟 Advertise your project 🚀

Facebook Opt 6.7B Qcqa Ub 16 Best For KV Cache Parameters and Internals

Model Type 
text generation
Additional Notes 
This version uses a modified Multi-Head Attention (MHA) architecture with 6 groups of KV heads per layer, each sharing mean-pooled K/V values for the same group instead of the original 12 separate KV heads.
Supported Languages 
en (primary)
LLM NameFacebook Opt 6.7B Qcqa Ub 16 Best For KV Cache
Repository 🤗https://huggingface.co/saarvajanik/facebook-opt-6.7b-qcqa-ub-16-best-for-KV-cache 
Model Size6.7b
Required VRAM27.4 GB
Updated2026-04-04
Maintainersaarvajanik
Model Typeopt
Model Files  10.0 GB: 1-of-3   9.9 GB: 2-of-3   7.5 GB: 3-of-3
Supported Languagesen
Model ArchitectureOPTForCausalLM
Licensemit
Context Length2048
Model Max Length2048
Transformers Version4.23.1
Tokenizer ClassGPT2Tokenizer
Beginning of Sentence Token</s>
End of Sentence Token</s>
Unk Token</s>
Vocabulary Size50272
Torch Data Typefloat32
Activation Functionrelu
Errorsreplace

Best Alternatives to Facebook Opt 6.7B Qcqa Ub 16 Best For KV Cache

Best Alternatives
Context / RAM
Downloads
Likes
....7B Qcqa Ub 16 Best For Q Loss2K / 27.4 GB5830
Galactica 6.7B ReFT GSM8k2K / 13.8 GB7150
...7B Gqa Ub 16 Best For KV Cache2K / 27.4 GB2280
Galactica 6.7B EssayWriter2K / 13.4 GB7565
Opt Flan Iml 6.7B2K / 26.6 GB9431
...lactica 6.7B Evol Instruct 70K2K / 27.4 GB42819
Galactica 6.7B Finetuned2K / 27.4 GB263934
OPT 6.7B Nerybus Mix2K / 13.7 GB89820
Opt 6.7B2K / 13.4 GB30449118
OPT 6.7B Erebus2K / 13.3 GB209498
Note: green Score (e.g. "73.2") means that the model is better than saarvajanik/facebook-opt-6.7b-qcqa-ub-16-best-for-KV-cache.

Rank the Facebook Opt 6.7B Qcqa Ub 16 Best For KV Cache 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