Llama 3 was released April 18, 2024 in 8B and 70B sizes. Trained on 15T tokens (7x Llama 2). The 70B model scored 82% on MMLU and 81.7% on HumanEval, approaching GPT-4 performance. Used a Grouped Query Attention (GQA) architecture and a 128K token vocabulary.
Benchmarks4
Knowledge
1 benchmarkReasoning
1 benchmarkMath
1 benchmarkCoding
1 benchmarkPercentile among tested models:Top 25%50-75%25-50%Bottom 25%
Llama Family5
| Model | Tier | Released | Input $/MTok |
|---|---|---|---|
| Llama 4 Scout | — | 2025-04-05 | — |
| Llama 4 Maverick | — | 2025-04-05 | — |
| Llama 3.3 | — | 2024-12-06 | — |
| Llama 3.1 | — | 2024-07-23 | — |
| Llama 2 | — | 2023-07-18 | — |
Details
Model FamilyLlama
Generation3
Release Date2024-04-18
Parameters70B
Context Window8K tokens
Open WeightYes
Modalitytext
Capabilities1
tool-use
Sources1
Tags
llamametaopen-weight