Meet Llama 3.1
The open source AI model you can fine-tune, distill and deploy anywhere. Our latest instruction-tuned model is available in 8B, 70B and 405B versions.
Llama 3.1 models
Documentation Hub
Key capabilities
Start building more advanced use cases, leveraging our resources.
Tool use
Upload a dataset and analyze it. Prompt model to plot graphs and fetch market data.
Make Llama your own
Using our open ecosystem, build faster with a selection of differentiated product offerings to support your use cases.
Inference
Choose from real-time inference or batch inference services. Download model weights to further optimize cost per token.
Fine-tune, Distill & Deploy
Adapt for your application, improve with synthetic data and deploy on-prem or in the cloud.
RAG & Tool Use
Use Llama system components and extend the model using zero shot tool use and RAG to build agentic behaviors.
Synthetic Data Generation
Leverage 405B high quality data to improve specialized models for specific use cases.Quick start with partners
Features for
405B models
405B models
Real-time inference
Batch inference
Fine-tuning
Model evaluation
RAG
Continual pre-training
Safety guardrails
Synthetic data generation
Distillation recipe
Model evaluations
As measured on over 150 benchmark datasets that span a wide range of languages and extensive human evaluations.
Llama 3.1 Performance
Benchmarks
Category
Benchmark
General
MMLU
(CoT)
MMLU PRO
(5-shot, CoT)
IFEval
Code
HumanEval
(0-shot)
MBPP EvalPlus
(base) (0-shot)
Math
GSM8K
(8-shot, CoT)
MATH
(0-shot, CoT)
Reasoning
ARC Challenge
(0-shot)
GPQA
(0-shot, CoT)
Tool use
API-Bank
(0-shot)
BFCL
Gorilla Benchmark API Bench
Nexus
(0-shot)
Multilingual
Multilingual MGSM
Llama 3.1
8B
73.0
48.3
80.4
72.6
72.8
84.5
51.9
83.4
32.8
82.6
76.1
8.2
38.5
68.9
Llama 3
8B - April
65.3
45.5
76.8
60.4
70.6
80.6
29.1
82.4
34.6
48.3
60.3
1.7
18.1
-
Llama 3.1
70B
86.0
66.4
87.5
80.5
86.0