LLM EXPLORER 63,407 MODELS INDEXED

Snorkel Mistral PairRM DPO by snorkelai

By snorkelai · 76 downloads

Snorkel Mistral PairRM DPO is an open-source language model by snorkelai. Features: LLM, VRAM: 14.4GB, Context: 32K, License: apache-2.0, LLM Explorer Score: 0.12, Arc: 66, HellaSwag: 85.6, MMLU: 60.9, GSM8K: 36.2.

  Arxiv:2305.18290   Arxiv:2306.02561   Arxiv:2312.11456   Arxiv:2401.10020   Conversational Dataset:snorkelai/snorkel-mist...   Endpoints compatible   Mistral   Pytorch   Region:us   Sharded

Snorkel Mistral PairRM DPO Parameters and Internals

Model Type 
text-generation
Use Cases 
Limitations:
The model is a quick demonstration and does not have any moderation mechanisms.
Additional Notes 
For enterprise use cases, additional fine-tuning and alignment are necessary. Interested parties can contact Snorkel AI for specialized reward models.
Training Details 
Data Sources:
snorkelai/Snorkel-Mistral-PairRM-DPO-Dataset, UltraFeedback
Methodology:
1. Generate five response variations for each prompt from a subset of 20,000 using the LLM - to start, we used [Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2). 2. Apply [PairRM](https://huggingface.co/llm-blender/PairRM) for response reranking. 3. Update the LLM by applying Direct Preference Optimization (DPO) on the top (chosen) and bottom (rejected) responses. 4. Use this LLM as the base model for the next iteration, repeating three times in total.
Input Output 
Input Format:
[INST] {prompt} [/INST]
Accepted Modalities:
text
Performance Tips:
The model is designed for initial trials and may take time initially to activate on Hugging Face endpoint.
Release Notes 
Version:
GGUF
Notes:
Model version available from community members.
LLM NameSnorkel Mistral PairRM DPO
Repository πŸ€—https://huggingface.co/snorkelai/Snorkel-Mistral-PairRM-DPO 
Required VRAM14.4 GB
Updated2026-07-30
Maintainersnorkelai
Model Typemistral
Model Files  9.9 GB: 1-of-2   4.5 GB: 2-of-2   0.0 GB
Model ArchitectureMistralForCausalLM
Licenseapache-2.0
Context Length32768
Model Max Length32768
Transformers Version4.34.0
Tokenizer ClassLlamaTokenizer
Padding Token</s>
Vocabulary Size32000
Torch Data Typebfloat16

Quantized Models of the Snorkel Mistral PairRM DPO

Model
Likes
Downloads
VRAM
...dle Snorkel Mistral PairRM DPO018114 GB

Best Alternatives to Snorkel Mistral PairRM DPO

Best Alternatives
Context / RAM
Downloads
Likes
Krutrim 2 Instruct1000K / 49.3 GB86638
Ft V1 Violet1000K / 24.5 GB50
Tiny Random MistralForCausalLM128K / 0 GB1541
Winterreise M732K / 14.4 GB00
Frostwind V2.1 M732K / 14.4 GB00
...ydaz Web AI Reasoner BaseModel32K / 14.4 GB01
MistralLite32K / 14.4 GB61777430
MistralLite32K / 14.4 GB2373437
Mixtral AI Cyber Child32K / 14.5 GB141
Kheops Textbook Immo232K / 14.5 GB110
Note: green Score (e.g. "73.2") means that the model is better than snorkelai/Snorkel-Mistral-PairRM-DPO.