Skip to main content
Language model

Llama 3.1 70B Instruct

Meta

Released
July 2024

Llama 3.1 70B Instruct tunes the 70-billion-parameter version of Llama 3.1 70B for multilingual dialogue. Meta combines supervised fine-tuning with reinforcement learning from human feedback.

Compared with Llama 3.1 8B Instruct, the parameter count grows while context remains at 128K tokens. Both versions support tool calls through chat templates and eight languages. The weights use the Llama 3.1 Community License, with a December 2023 knowledge cutoff.

Specifications and access

SpecificationValue and source
Weights
Open weightsSource
Access
APISource
Parameters
70 billionSource
Context window
131,072 tokensSource
Knowledge Cutoff
December 2023Source
License
Llama 3.1 CommunitySource
Inputs
TextSource
Outputs
TextSource
Tool calling
YesSource