Locutusque/hyperion-medium-preview

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:8kPublished:Feb 24, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

Locutusque/hyperion-medium-preview is a 7 billion parameter language model, fine-tuned by M4-ai on the Hyperion dataset, building upon the Mistral-7B-v0.1 base. This model specializes in advanced reasoning across scientific domains, excelling in complex question answering, conversational AI, code generation, medical text comprehension, mathematical reasoning, and logical reasoning. It is particularly optimized for handling intricate inquiries and instructions within technical and scientific contexts, making it suitable for AI-driven tutoring and domain-specific information retrieval.

Loading preview...

Model Overview

Locutusque/hyperion-medium-preview is a 7 billion parameter language model developed by M4-ai, fine-tuned from the mistralai/Mistral-7B-v0.1 base. It is specifically trained on the Hyperion dataset, which is rich in diverse and complex information spanning programming, medical texts, mathematical problems, and various reasoning tasks. This model is designed for advanced reasoning across multiple scientific domains.

Key Capabilities

  • Complex Question Answering: Handles intricate inquiries across scientific and technical subjects.
  • Conversational AI: Supports conversational understanding with a focus on technical and scientific reasoning.
  • Code Generation: Capable of generating and understanding complex programming contexts.
  • Domain-Specific Comprehension: Excels in medical text comprehension, mathematical reasoning, and logical reasoning.

Intended Use Cases

  • AI-driven Tutoring Systems: Ideal for science, medicine, mathematics, and computer science education.
  • Assistive Tools: Provides fast and accurate domain-specific information retrieval for professionals.
  • Technical Platforms: Suitable for applications requiring advanced conversational AI in technical and scientific fields.

Performance Highlights

Evaluations on the Open LLM Leaderboard show an average score of 61.67. Notable scores include 83.67 on HellaSwag (10-Shot) and 63.73 on MMLU (5-Shot), demonstrating its strong reasoning and comprehension abilities. The model is released under the Apache-2.0 license.

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