LLM EXPLORER 59,420 MODELS INDEXED

L3 SnowStorm 4x8B 6.5bpw H8 EXL2 by JayhC

By JayhC · 5 downloads
HF Repo ↗ Copy URL

L3 SnowStorm 4x8B 6.5bpw H8 EXL2 is an open-source language model by JayhC. Features: 8b LLM, VRAM: 21GB, Context: 8K, License: llama3, MoE, Quantized, LLM Explorer Score: 0.12.

  Conversational   En   Endpoints compatible   Exl2   Mixtral   Moe   Quantized   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/JayhC/L3_SnowStorm_4x8B-6.5bpw-h8-exl2 

L3 SnowStorm 4x8B 6.5bpw H8 EXL2 Parameters and Internals

Model Type 
moe
Additional Notes 
Quantization using exllamav2 0.0.21 and default calibration dataset. Aimed at RP/ERP tasks with a goal to outperform Mixtral 8x7B.
LLM NameL3 SnowStorm 4x8B 6.5bpw H8 EXL2
Repository 🤗https://huggingface.co/JayhC/L3_SnowStorm_4x8B-6.5bpw-h8-exl2 
Model Size8b
Required VRAM21 GB
Updated2026-06-30
MaintainerJayhC
Model Typemixtral
Model Files  8.6 GB: 1-of-3   8.6 GB: 2-of-3   3.8 GB: 3-of-3
Supported Languagesen
Quantization Typeexl2
Model ArchitectureMixtralForCausalLM
Licensellama3
Context Length8192
Model Max Length8192
Transformers Version4.41.0
Tokenizer ClassPreTrainedTokenizerFast
Padding Token<|begin_of_text|>
Vocabulary Size128256
Torch Data Typebfloat16

Best Alternatives to L3 SnowStorm 4x8B 6.5bpw H8 EXL2

Best Alternatives
Context / RAM
Downloads
Likes
...Storm V1.15 4x8B B 8 0bpw EXL28K / 25.2 GB60
L3 SnowStorm 4x8B 8bpw H8 EXL28K / 25.2 GB61
...oquy V2 4x8B Test 8bpw H8 EXL28K / 24.9 GB51
...liloquy V1.5 4x8B 8bpw H8 EXL28K / 25.3 GB102
...loquy V1.5 4x8B 6.5bpw H8 EXL28K / 21 GB62
...oquy 4x8B Bpw600 H6 EXL2 Rpcal8K / 19.4 GB52
...oquy 4x8B Bpw500 H6 EXL2 Rpcal8K / 16.4 GB41
...ticSoliloquy 4x8B 8bpw H8 EXL28K / 25.3 GB63
...oquy 4x8B Bpw800 H8 EXL2 Rpcal8K / 25.3 GB01
...ticSoliloquy 4x8B 6bpw H6 EXL28K / 19.4 GB23
Note: green Score (e.g. "73.2") means that the model is better than JayhC/L3_SnowStorm_4x8B-6.5bpw-h8-exl2.