Atlas

Models

← All models

Llama 3 8B

Meta · 2024-04-18 · 8B parameters

Llama 3 8B is Meta's 8-billion-parameter pretrained base language model from the Llama 3 family, released in April 2024. It is a dense decoder-only Transformer trained on a large multilingual corpus with an 8K-token context window, intended as the foundation checkpoint for downstream fine-tuning including the Llama 3 8B Instruct model.

Benchmark scores

BenchmarkScore
Capability110.3
Epoch Capabilities Index (ECI)116
ForecastBench Tournament Overall Brier Index52.9
MedMCQA (Open Medical-LLM)57.5
MedQA 4-options (Open Medical-LLM)59.7
MMLU Anatomy (Open Medical-LLM)68.9
MMLU Clinical Knowledge (Open Medical-LLM)74.7
MMLU College Biology (Open Medical-LLM)78.5
MMLU College Medicine (Open Medical-LLM)61.8
MMLU Medical Genetics (Open Medical-LLM)83.0
MMLU Professional Medicine (Open Medical-LLM)70.2
PubMedQA (Open Medical-LLM)74.8
WinoGrande75.7
Loading Atlas data…