arcee-ai/mistral-sliced

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7BQuant:FP8Context Size:4kTool Calling:SupportedPublished:Mar 13, 2024License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The arcee-ai/mistral-sliced model is a 916 million parameter variant derived from the Mistral-7B-v0.1 base model. Created using mergekit, this model is specifically designed for testing purposes rather than generating coherent text. It represents a 'sliced' version, utilizing a subset of layers from its larger Mistral counterpart.

Loading preview...

Model Overview

The arcee-ai/mistral-sliced model is a specialized variant of the mistralai/Mistral-7B-v0.1 base model, created through a 'slicing' process using mergekit. This process extracts a specific range of layers from the original model, resulting in a significantly smaller model with 916,484,096 parameters.

Key Characteristics

  • Derived from Mistral-7B-v0.1: Built upon the architecture of the well-known Mistral 7B model.
  • Parameter Count: Contains approximately 916 million parameters, making it much smaller than its 7B base.
  • Slicing Method: Utilizes mergekit to select layers 0 through 3 from the original Mistral-7B-v0.1 model.
  • Intended Use: Explicitly designed for testing and experimental purposes.

Important Note

This model is not intended for generating coherent text. Its primary function is for testing and understanding the effects of model slicing, rather than for practical natural language generation applications.