mylesgoose/Llama-3.2-1B-Instruct-abliterated3

Hugging Face
TEXT GENERATIONPricing:Input $0.108 / Output $0.804Concurrent Unit Cost:1Model Size:1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Oct 1, 2024License:llama3.2Architecture:Transformer Featherless Exclusive Warm

mylesgoose/Llama-3.2-1B-Instruct-abliterated3 is a 1.23 billion parameter instruction-tuned, multilingual large language model developed by Meta, part of the Llama 3.2 family. Optimized for multilingual dialogue, agentic retrieval, and summarization tasks, it features a 32768-token context length and utilizes an optimized transformer architecture with Grouped-Query Attention. This model is designed for commercial and research use, outperforming many open-source and closed chat models on common benchmarks.

Loading preview...

Model Overview

mylesgoose/Llama-3.2-1B-Instruct-abliterated3 is a 1.23 billion parameter instruction-tuned model from Meta's Llama 3.2 collection, optimized for multilingual dialogue and agentic applications. It leverages an optimized transformer architecture with Grouped-Query Attention (GQA) and has a context length of 32768 tokens. The model was pretrained on up to 9 trillion tokens of publicly available data, with knowledge distillation from larger Llama 3.1 models, and fine-tuned using SFT, Rejection Sampling, and DPO for alignment with human preferences.

Key Capabilities

  • Multilingual Support: Officially supports English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai, with training on a broader set of languages.
  • Dialogue and Agentic Tasks: Optimized for assistant-like chat, knowledge retrieval, summarization, and mobile AI-powered writing assistants.
  • Performance: Outperforms many open-source and closed chat models on common industry benchmarks, including strong results on MMLU, GSM8K, and multilingual MMLU.
  • Long Context: Demonstrates capabilities in long-context tasks with a 32k context window.

Good For

  • Commercial and Research Use: Intended for a wide range of applications in both commercial products and academic research.
  • Constrained Environments: The 1B and 3B models are specifically designed for deployment in highly constrained environments, such as mobile devices.
  • Developers: Provides a valuable resource for studying safety fine-tuning and offers a readily available, safe, and powerful model for various applications, reducing developer workload for deploying safe AI systems.

Finetunes
4 models