LLM EXPLORER 59,420 MODELS INDEXED

Rogue Rose 103B V0.2 GPTQ by TheBloke

By TheBloke · 3 downloads
HF Repo ↗ Copy URL

Rogue Rose 103B V0.2 GPTQ is an open-source language model by TheBloke. Features: 103b LLM, VRAM: 52.5GB, Context: 4K, License: llama2, Quantized, LLM Explorer Score: 0.1.

  Arxiv:2307.11760   4-bit Base model:quantized:sophosymp... Base model:sophosympatheia/rog...   En   Gptq   Llama   Quantized   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/TheBloke/Rogue-Rose-103b-v0.2-GPTQ 

Rogue Rose 103B V0.2 GPTQ Parameters and Internals

Model Type 
llama
Use Cases 
Limitations:
May have limitations in scene logic
Additional Notes 
Recommended sampler settings from Reddit guide. Experiment with prompt and system prompt for better results.
Supported Languages 
en (English)
Training Details 
Context Length:
4096
Hardware Used:
Massed Compute
Model Architecture:
Frankenmerge of two custom 70b merges with 120 layers.
Input Output 
Input Format:
Vicuna-Short: 'You are a helpful AI assistant. USER: {prompt} ASSISTANT: '
Accepted Modalities:
text
Output Format:
Varies depending on usage and template
Performance Tips:
Try using the new Min-P sampler method. Recommended settings: temperature at high levels, repetition penalty higher than normal, presence penalty higher than normal.
Release Notes 
Version:
3.2 bpw
Notes:
Fits within 48 GB of VRAM at 8192 context.
Version:
3.5 bpw (PENDING)
Notes:
Barely fits within 48 GB of VRAM at ~4096 context using the 8-bit cache setting.
LLM NameRogue Rose 103B V0.2 GPTQ
Repository 🤗https://huggingface.co/TheBloke/Rogue-Rose-103b-v0.2-GPTQ 
Model NameRogue Rose 103B v0.2
Model CreatorSophosympatheia
Base Model(s)  sophosympatheia/Rogue-Rose-103b-v0.2   sophosympatheia/Rogue-Rose-103b-v0.2
Model Size103b
Required VRAM52.5 GB
Updated2026-08-04
MaintainerTheBloke
Model Typellama
Model Files  49.0 GB: 1-of-2   3.5 GB: 2-of-2
Supported Languagesen
GPTQ QuantizationYes
Quantization Typegptq
Model ArchitectureLlamaForCausalLM
Licensellama2
Context Length4096
Model Max Length4096
Transformers Version4.37.0.dev0
Tokenizer ClassLlamaTokenizer
Padding Token</s>
Vocabulary Size32000
Torch Data Typefloat16

Best Alternatives to Rogue Rose 103B V0.2 GPTQ

Best Alternatives
Context / RAM
Downloads
Likes
Aurora Nights 103B V1.0 GPTQ4K / 52.5 GB83
Solus 103B L2 GPTQ4K / 52.6 GB31
Lila 103B L2 GPTQ4K / 52.6 GB81
... Dark Miqu 103B 5.0bpw H8 EXL231K / 65.1 GB100
...qu 103B V1.5 EXL2 5.0bpw Rpcal31K / 64.9 GB91
... Miqu 103B V1.5 3.5bpw H6 EXL231K / 45.6 GB21
...qu 103B V1.5 EXL2 4.0bpw Rpcal31K / 52.1 GB31
...qu 103B V1.5 EXL2 3.0bpw Rpcal31K / 39.3 GB21
...ght Miqu 103B V1.0 EXL2 4.5bpw31K / 58.7 GB71
...ght Miqu 103B V1.0 EXL2 4.0bpw31K / 52.2 GB61
Note: green Score (e.g. "73.2") means that the model is better than TheBloke/Rogue-Rose-103b-v0.2-GPTQ.