LLM EXPLORER 59,516 MODELS INDEXED

L3.2 8X4B MoE V2 Dark Champion Inst 21B Uncen Ablit by DavidAU

By DavidAU · 87 downloads
HF Repo ↗ Copy URL

L3.2 8X4B MoE V2 Dark Champion Inst 21B Uncen Ablit is an open-source language model by DavidAU. Features: 20.9b LLM, VRAM: 83.3GB, Context: 128K, MoE, Instruction-Based, LLM Explorer Score: 0.17.

  128k context   8 experts   8 x 4b   Autotrain compatible   Brainstorm 5x   Conversational   Endpoints compatible   Fine tune   Instruct   Llama 3.2   Mixtral   Mixture of experts   Moe   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/DavidAU/L3.2-8X4B-MOE-V2-Dark-Champion-Inst-21B-uncen-ablit 

L3.2 8X4B MoE V2 Dark Champion Inst 21B Uncen Ablit Parameters and Internals

LLM NameL3.2 8X4B MoE V2 Dark Champion Inst 21B Uncen Ablit
Repository 🤗https://huggingface.co/DavidAU/L3.2-8X4B-MOE-V2-Dark-Champion-Inst-21B-uncen-ablit 
Model Size20.9b
Required VRAM83.3 GB
Updated2025-09-23
MaintainerDavidAU
Model Typemixtral
Instruction-BasedYes
Model Files  5.0 GB: 1-of-17   4.9 GB: 2-of-17   4.9 GB: 3-of-17   4.9 GB: 4-of-17   4.9 GB: 5-of-17   4.9 GB: 6-of-17   4.9 GB: 7-of-17   4.9 GB: 8-of-17   4.9 GB: 9-of-17   5.0 GB: 10-of-17   5.0 GB: 11-of-17   4.9 GB: 12-of-17   4.9 GB: 13-of-17   4.9 GB: 14-of-17   4.9 GB: 15-of-17   4.9 GB: 16-of-17   4.6 GB: 17-of-17
Model ArchitectureMixtralForCausalLM
Context Length131072
Model Max Length131072
Transformers Version4.46.2
Tokenizer ClassPreTrainedTokenizerFast
Padding Token<|begin_of_text|>
Vocabulary Size128256
Torch Data Typefloat32

Best Alternatives to L3.2 8X4B MoE V2 Dark Champion Inst 21B Uncen Ablit

Best Alternatives
Context / RAM
Downloads
Likes
...nsored Abliterated 21B Q 6 MLX128K / 18.3 GB7143
Note: green Score (e.g. "73.2") means that the model is better than DavidAU/L3.2-8X4B-MOE-V2-Dark-Champion-Inst-21B-uncen-ablit.