Llama 3.1 70B

by Meta · llama-3 family

70B

parameters

text-generation code-generation reasoning multilingual tool-use math creative-writing summarization

Llama 3.1 70B is Meta's large-scale open model delivering near-frontier performance across a wide range of tasks. It excels at complex reasoning, coding, mathematical problem solving, and creative writing. With 70 billion parameters, this model requires significant hardware but rewards users with exceptional quality. It supports 128K context, native tool use, and multilingual capabilities across eight languages.

Quick Start with Ollama

ollama run 70b-instruct-q4_K_M
Creator Meta
Parameters 70B
Architecture transformer-decoder
Context Length 128K tokens
License Llama 3.1 Community License
Released Jul 23, 2024
Ollama llama3.1:70b

Quantization Options

Format File Size VRAM Required Quality Ollama Tag
Q4_K_M recommended 34.9 GB 43.5 GB
70b-instruct-q4_K_M
Q5_K_M 40.8 GB 50.5 GB
70b-instruct-q5_K_M
Q8_0 64.8 GB 72 GB
70b-instruct-q8_0

Compatible Hardware for Q4_K_M

Showing compatibility for the recommended quantization (Q4_K_M, 43.5 GB VRAM).

Compatible Hardware

Hardware VRAM Type Fit
Mac Pro M2 Ultra 192GB 192 GB mac Runs
Mac Studio M4 Ultra 192GB 192 GB mac Runs
Mac Studio M4 Max 128GB 128 GB mac Runs
MacBook Pro M4 Max 128GB 128 GB mac Runs
Mac Studio M4 Max 64GB 64 GB mac Runs
MacBook Pro M4 Max 64GB 64 GB mac Runs
Mac mini M4 Pro 48GB 48 GB mac Runs (tight)
MacBook Pro M4 Max 48GB 48 GB mac Runs (tight)
MacBook Pro M4 Pro 48GB 48 GB mac Runs (tight)
NVIDIA GeForce RTX 5090 32 GB gpu CPU Offload
Mac mini M4 32GB 32 GB mac CPU Offload
25 hardware device(s) cannot run this model configuration.

Benchmark Scores

83.6
mmlu