L

Llama 3.2 1B Instruct

Developed by meta-llama
Llama 3.2 is a multilingual large language model series developed by Meta, including 1B and 3B scale pre-trained and instruction-tuned generative models, optimized for multilingual dialogue scenarios, supporting intelligent retrieval and summarization tasks.
Downloads 2.4M
Release Time : 9/18/2024

Model Overview

The Llama 3.2 plain-text instruction-tuned model is optimized for multilingual dialogue scenarios, including intelligent retrieval and summarization tasks. In common industry benchmarks, it outperforms many open-source and closed-source chat models.

Model Features

Multilingual Support
Supports 8 official languages, with a broader training language range, allowing fine-tuning for more languages.
Long Context Handling
Supports 128k context length, suitable for processing long documents and complex dialogue scenarios.
Efficient Inference
Uses Grouped Query Attention (GQA) to improve inference scalability, with quantized versions suitable for resource-limited devices.
Knowledge Distillation
Integrates logits from Llama 3.1 8B and 70B models as token-level targets, recovering pruned performance through knowledge distillation.

Model Capabilities

Multilingual Text Generation
Intelligent Retrieval
Text Summarization
Dialogue Systems
Query/Prompt Rewriting

Use Cases

Intelligent Assistants
Multilingual Chatbot
Build a dialogue assistant supporting multiple languages for global users.
Excellent performance in officially supported languages
Mobile AI Writing Assistant
Provides real-time writing suggestions and content generation on mobile devices.
Quantized versions suitable for resource-limited devices
Information Processing
Knowledge Retrieval
Quickly retrieve relevant information from large texts.
Good performance in reading comprehension benchmarks
Text Summarization
Generate concise summaries of long documents.
RougeL score of 16.8-19.2 in TLDR9+ tests
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase