huihui-ai/DeepSeekR1-QwQ-SkyT1-32B-Fusion-811

Hugging Face
TEXT GENERATIONConcurrency Cost:2Model Size:32BQuant:FP8Ctx Length:32kLicense:apache-2.0Architecture:Transformer0.0K Open Weights Warm

The huihui-ai/DeepSeekR1-QwQ-SkyT1-32B-Fusion-811 is a 32 billion parameter language model built on the Qwen 2.5 architecture. This model is a fusion of three distinct Qwen-based models: DeepSeek-R1-Distill-Qwen-32B, QwQ-32B-Preview, and Sky-T1-32B-Preview, blended in an 80:10:10 ratio. It is designed to combine the strengths of its constituent models, offering a usable and coherent output despite being a simple mix.

Loading preview...

Model Overview

huihui-ai/DeepSeekR1-QwQ-SkyT1-32B-Fusion-811 is a 32 billion parameter language model based on the Qwen 2.5 architecture. This model is a unique fusion of three distinct Qwen-based models, specifically:

  • huihui-ai/DeepSeek-R1-Distill-Qwen-32B-abliterated (contributing 80%)
  • huihui-ai/QwQ-32B-Preview-abliterated (contributing 10%)
  • huihui-ai/Sky-T1-32B-Preview-abliterated (contributing 10%)

Key Characteristics

This model aims to leverage the combined strengths of its constituent parts. Despite being a straightforward blend, the developers note that the model produces usable output without exhibiting gibberish, indicating a successful integration of the different model components. The project also explores different blending ratios (e.g., 70:15:15 and 60:20:20) to assess their impact on model performance.

Usage

For users interested in deploying this model, it is available for use with Ollama. A direct command is provided for easy integration:

ollama run huihui_ai/deepseekr1-qwq-skyt1-fusion

Popular Sampler Settings

Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.

temperature
top_p
top_k
frequency_penalty
presence_penalty
repetition_penalty
min_p