L

Llama 3.3 70B Instruct

Developed by meta-llama
Meta Llama 3.3 is a 70-billion-parameter multilingual large language model optimized for multilingual dialogue scenarios, supporting 8 languages and excelling in multiple benchmark tests.
Downloads 1.1M
Release Time : 11/26/2024

Model Overview

An autoregressive language model based on an optimized Transformer architecture, employing supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF) to align with human preferences, suitable for multilingual text generation and dialogue tasks.

Model Features

Multilingual support
Supports text generation and dialogue tasks in 8 languages
Long context processing
Supports a long context window of 128k tokens
Tool calling capability
Supports integration and calling of external tools
Quantized deployment
Supports 4/8-bit quantized deployment, reducing hardware requirements

Model Capabilities

Multilingual text generation
Instruction following
Tool calling
Code generation
Mathematical reasoning
Knowledge Q&A

Use Cases

Dialogue systems
Multilingual customer service assistant
Build intelligent customer service systems supporting multiple languages
Achieved a score of 86.0 on the MMLU benchmark
Content generation
Multilingual content creation
Generate multilingual marketing copy, articles, and other content
Education
Language learning assistant
Help learners practice multilingual conversation and writing
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase