agentlans/Llama3.1-SuperDeepFuse
Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kPublished:Jan 20, 2025License:llama3.1Architecture:Transformer0.0K Warm

agentlans/Llama3.1-SuperDeepFuse is an 8 billion parameter language model based on Llama 3.1, created by agentlans. This model is a merge of three high-performance distilled models, designed to enhance multi-task reasoning, instruction-following, and performance in mathematics and coding. It offers balanced performance across diverse tasks and is suitable for consumer GPU deployment.

Loading preview...

Llama3.1-SuperDeepFuse: Merged for Enhanced Reasoning

Llama3.1-SuperDeepFuse is an 8 billion parameter language model developed by agentlans, built upon the meta-llama/Llama-3.1-8B-Instruct base. This model distinguishes itself by merging three high-performance distilled models: arcee-ai/Llama-3.1-SuperNova-Lite, deepseek-ai/DeepSeek-R1-Distill-Llama-8B, and FuseAI/FuseChat-Llama-3.1-8B-Instruct, using the model_stock merge method.

Key Capabilities

  • Enhanced Multi-Task Reasoning: Designed to improve complex problem-solving across various domains.
  • Improved Mathematical and Coding Performance: Specifically targets better accuracy and utility in quantitative and programming tasks.
  • Multilingual Support: Offers capabilities across multiple languages.
  • Consumer GPU Deployment: Optimized for accessibility on standard hardware.

Performance Notes

The model maintains Llama 3.1's safety standards and aims for balanced performance. While still undergoing benchmarking, initial evaluations on the Open LLM Leaderboard show an Average score of 27.30%, with notable results in IFEval (77.62%) and MMLU-PRO (30.83%). It's important to note that, like all language models, it can produce misleading output, and results should be independently verified. Its capabilities are limited compared to larger model variants.

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