Yuma42/KangalKhan-Sapphire-7B

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Feb 15, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

KangalKhan-Sapphire-7B is a 7 billion parameter language model developed by Yuma42, created by merging argilla/CapybaraHermes-2.5-Mistral-7B and argilla/distilabeled-OpenHermes-2.5-Mistral-7B using slerp. This model, built on the Mistral architecture, features a 4096-token context length and demonstrates strong general-purpose performance across various benchmarks, including reasoning, common sense, and language understanding tasks. It is suitable for applications requiring robust conversational AI and text generation capabilities.

Loading preview...

KangalKhan-Sapphire-7B Overview

KangalKhan-Sapphire-7B is a 7 billion parameter language model developed by Yuma42. It is a product of merging two base models, argilla/CapybaraHermes-2.5-Mistral-7B and argilla/distilabeled-OpenHermes-2.5-Mistral-7B, utilizing the slerp merge method. This approach combines the strengths of its constituent models to offer enhanced performance.

Key Capabilities & Performance

Evaluated on the Open LLM Leaderboard, KangalKhan-Sapphire-7B achieves an average score of 68.52. Specific benchmark results include:

  • AI2 Reasoning Challenge (25-Shot): 66.30
  • HellaSwag (10-Shot): 85.34
  • MMLU (5-Shot): 63.32
  • TruthfulQA (0-shot): 56.09
  • Winogrande (5-shot): 78.14
  • GSM8k (5-shot): 61.94

These scores indicate strong performance across a range of tasks, including reasoning, common sense, and language understanding. The model is configured with a 4096-token context length and supports bfloat16 for efficient computation.

Recommended Use Cases

This model is well-suited for general-purpose applications requiring robust text generation and conversational AI. Its balanced performance across various benchmarks makes it a versatile choice for tasks such as:

  • Content creation
  • Chatbot development
  • Question answering
  • Reasoning-based tasks